A Systematic Review and Meta-Analysis of Physiologic Variables in Sheep Fetuses
The fetal sheep model has been widely used in fetal therapy research. However, there is a significant degree of variability among published normal values. Our study aimed to evaluate the literature available on normal values for hemodynamics, blood gases, and acid-base status in the sheep fetus and to determine the best possible estimation of such physiologic values. We conducted a systematic review with a comprehensive search of several databases. We included 189 articles in the database and over 2,800 sheep fetuses. Analysis revealed a mean umbilical blood flow of 202 mL/kg/min (95% CI: 182 to 223); mean arterial pCO2 of 49.8 mm Hg (95% CI: 49.2 to 50.3); mean arterial pO2 of 22.3 mm Hg (95% CI: 21.9 to 22.7); mean arterial pH of 7.35 (95% CI: 7.3487 to 7.3562); and mean arterial oxygen saturation of 59.8 (95% CI; 58 to 61.7). Our findings were punctuated by a high heterogeneity, for which we conducted several subanalyses. The results showed high heterogeneity and small study effect in the literature available and provided our best assessment of relevant variables on normal hemodynamics, blood gases, and acid-base status in the fetus after using strategies to mitigate the risk of bias present in the literature.
Introduction
Animal models play a crucial role in advancing medicine and human health. These models allow researchers to learn more about diseases and innovative treatment feasibility, safety, and effectiveness before human trials. This is particularly important in fetal therapy, in which the treatment of the fetus implies the exposure of the mother to potential side effects.
Since the 1960s fetal therapy has advanced tremendously, in part thanks to the use of animal models. 1,12,15,17,22 Key advancements include well-established procedural and surgical access such ultrasound guided needle procedures and endoscopic and open surgical approaches. There is now scientific evidence supporting in utero treatment of several conditions, such as twin-twin transfusion syndrome, myelomeningocele, and congenital diaphragmatic hernia. 2,9,28 The fetal lamb is the most widely used model for testing new interventions before its application in humans. Many studies have proven the developmental similarities between human and sheep fetuses. 6 Moreover, the ease of access to fetal lambs allows for better feasibility of the studies. 23
Although the fetal sheep model has been widely used in fetal therapy research, there is a significant degree of variability among published normal values. To demonstrate the safety and efficacy of a treatment in animal models, first, we must know the normal or physiologic ranges of the variables to be affected by the treatment. For this reason, our group set out to perform a systematic review, use the most reliable information to determine normal ranges for fetal variables, and perform meta-analyses when applicable.
Methods
Study protocol.
This systematic review was conducted according to a prespecified protocol registered at PROSPERO for animal studies (CRD42021295771; https://www.crd.york.ac.uk/PROSPERO). This review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. No language restrictions were applied.
Data sources and search strategy.
A comprehensive search of several databases from each database’s inception to October 14th, 2021, English language, was conducted. The databases included Ovid MEDLINE(R) and Epub Ahead of Print, In-Process and Other Non-Indexed Citations, and Daily, Ovid EMBASE, Ovid Cochrane Central Register of Controlled Trials, Ovid Cochrane Database of Systematic Reviews, and Scopus. The search strategy was designed and conducted by an experienced librarian with input from the study’s principal investigator. Controlled vocabulary supplemented with keywords was used to search for studies of physiologic parameters of sheep fetuses. The actual strategy listing all search terms used and how they are combined is available in Figure 1.
Citation: Comparative Medicine 74, 5; 10.30802/AALAS-CM-24-033
Study selection: inclusion and exclusion criteria.
Two independent investigators (ATAT and SWF) screened articles using predefined inclusion and exclusion criteria. To be eligible for inclusion, papers had to provide 1) physiologic or baseline parameters (the only intervention allowed was surgery for probe placement and catheterization), 2) well-defined gestational age and subject number, and 3) include at least one of umbilical blood-flow, blood gas analysis (including pH, oxygen saturation [SatO2], pCO2, and pO2), or blood volume (fetal and/or placental). By physiologic or baseline parameters, we mean the values expected for a fetus without known pathology. Studies had to have more than one fetus as part of the publication to be included.
Study selection was performed as follows: by title, by abstract, and then by full text. Two researchers performed selection (ATAT and SWF), if no agreement was achieved in a discrepancy, a third author would be consulted.
A preliminary search generated 1,327 results dating back to 1945. Due to the high volume of results and expected changes in laboratory tests and flow assessment strategies, we opted to limit our database to publications on or after the year 2000 (n = 454).
Data extraction.
Blood gas.
Blood gas analysis was included if at least 2 of 4 parameters (pH, SatO2, pCO2, and pO2) were described, as well as the origin of the analyzed blood. Results were included in millimeters per mercury [if values were provided in kPa they were converted to mm Hg before addition to the database]. The artery that was monitored in each study varied (see Tables S1 to S3); therefore, we grouped these arteries into preductal and postductal and performed a subanalysis based on and comparing groups.
Umbilical blood flow.
Umbilical blood flow is reported (either in mL/kg/min or mL/min.) The technical details of how the measurement was obtained were also abstracted due to possible variations between methods.
Blood volume.
Blood volume in the fetus, placenta, or fetal-placenta system was included (registered in mL or mL/kg and converted if necessary).
Statistical analysis.
To establish a pooled analysis of the included studies, reported medians were converted to means using previously published methods. 31 The adjusted mean and 95% CIs were estimated from studies on outcomes of interest using hierarchical generalized linear models where the study was the random effect. This method accounts for variation among studies. Study heterogeneity was tested using the Cochran Q test statistic which is based on a χ2 distribution. The magnitude of the heterogeneity between studies is quantified by the I 2 statistic. 13 Heterogeneity was assessed by sensitivity analysis of key clinical characteristics of gestational age (less than 120, 120 to 129, 130 to 139, and > 140 d) and location of the blood draw in relation to the aortic arch (we divided this in preductal brachial, carotid, axillary, ascending aorta, aortic arch, postductal abdominal and thoracic aorta, femoral, pedal, and iliac and missing when location was not provided). Tests of small study effects were evaluated using the Egger test 22 and funnel plots for each outcome are presented. Analyses were performed using R (R Core Team [2021]) and the meta package. 5 Significance testing was based on a 2-sided test and results were considered statistically significant at the P < 0.05 cutoff.
Results
Our search amounted to a total of 454 articles. After initial screening by title and abstract, 341 full texts were reviewed, and 188 articles were included in the database. Figure 2 depicts the PRISMA flow diagram. Our main reason for exclusion was the absence of our target population (sheep fetuses) or the variables of interest. Records with missing data, reporting only data after interventions or graphs without describing the data, as well as reviews, were excluded.
Citation: Comparative Medicine 74, 5; 10.30802/AALAS-CM-24-033
In total, 2,838 observations were found within the 189 included articles. The average number of observations per article is 15 with an SD of 13.3 and a median of 12 with a range of 4 to 110.
Unfortunately, very few studies included venous blood gas or fetal blood volume data, limiting any relevant analysis.
Study characteristics.
The average gestational age for the articles included was 122 days (SD = 11.6 d) with a median of 125 d and a range of 90 to 140 d. A majority (n = 102) of the articles had study designs with more than one arm, with a median of 2 arms per study and a range of 2 to 7 arms. Among the articles that included blood gas analysis, 142 provided details on the source of blood collection (68 preductal; 74 postductal), 25 publications provided insufficient details, and 22 did not provide any information. No studies contained all outcomes of interest (Table S1), and the most reported outcomes were arterial pH (n = 169), pCO2 (n = 167), and pO2 (n = 162). Table 1 depicts a summary of the means obtained by random effect models for each variable.
| Mean | 95% CI | |
|---|---|---|
| Arterial oxygen saturation | 59.8% | 58–61.7 |
| Arterial pCO2 | 49.8 mm Hg | 49.2–50.3 |
| Arterial pO2 | 22.3 mm Hg | 21.9–22.7 |
| Arterial pH | 7.35 | 7.34–7.35 |
| Umbilical blood flow | 202.3 mL/kg/min | 181.9–222.7 |
Oxygen saturation.
A total of 71 studies were included in the blood SatO2 analysis. The mean obtained by random effects models was 59.8 (95% CI: 58 to 61.7).
Heterogeneity, the measure of variability among studies, was noted with an I 2 of 99.7%, where 75% to 100% is categorized as considerable heterogeneity. 14 Subgroup analysis of blood source considering preductal, postductal, or missing, found differences between preductal (65.3; 95% CI: 63.5 to 67.1), postductal (54.2; 95% CI: 51.4 to 57.1) and mean SatO2 (P < 0.0001). An additional subgroup analysis according to gestational age found significant differences between those greater than 140 d gestational age (65.3; 95% CI: 64.5 to 66.1) and the intervals of 120 to 129 d gestational age (57.5; 95% CI: 54.8 to 60.2), and 130 to 139 d gestational age (60.8; 95% CI: 57.9 to 63.8). The funnel plot (Figure 3) showed most studies falling outside of the 95% CI, with many to the left of the mean, which could represent bias due to reporting error, underreporting findings, or small study size effect. In this case, the Egger test for small study effects was significant (P < 0.0001), supporting the idea of small study effects.
Citation: Comparative Medicine 74, 5; 10.30802/AALAS-CM-24-033
Arterial pCO2.
A total of 167 studies were included in the pCO2 analysis. The mean value obtained by random effect models was 49.8 mm Hg (95% CI: 49.2 to 50.3). Heterogeneity was statistically identified with an I 2 of 99.4%. Subgroup analysis of blood source considering preductal (49.1; 95% CI: 48.3 to 49.9) and postductal (51.2; 95% CI: 50.3 to 52.1) found a significant difference (P = 0.0001). An additional subgroup analysis according to gestational age found that those with greater than 140 d gestational age (57.8; 95% CI: 51.8 to 63.9) had higher pCO2 scores than those with less than 120 d gestational age (49; 95% CI: 48.2 to 49.7) and those within 120 to 130 d gestational age (49.6; 95% CI: 48.2 to 52.3). No significant difference was noted with the 130 to 139 d gestational age group.
The funnel plot (Figure 4) showed a cluster of responses at the top of the funnel, with mean estimates between 40 and 60 for most studies. The pattern of responses was missing at the bottom of the funnel, suggesting that estimates about the means have a narrow range of values. The Egger test for small study effects was found to be significant (P = 0.006), supporting the idea of small study effects.
Citation: Comparative Medicine 74, 5; 10.30802/AALAS-CM-24-033
Arterial pO2.
A total of 162 publications were included in arterial pO2 analysis. The mean obtained by random effect models was 22.3 (95% CI: 21.9 to 22.7). When assessing for heterogeneity a high degree of variability was noted (I 2 of 98.8%). Subgroup analysis of blood source considering preductal, postductal, or missing found a slight difference between preductal (22.87; 95% CI: 22.43 to 23.32) and postductal (21.49; 95% CI: 20.97 to 22.01) mean pO2 (P = 0.0003). An additional subgroup analysis according to gestational age did not find any significant differences.
The funnel plot (Figure 5) showed a cluster of responses at the top of the funnel, with mean estimates between 10 and 30 for most studies. The pattern of responses was missing at the bottom of the funnel, suggesting that estimates about the means have a narrow range of values. The egger test for small study effect was not significant (P = 0.59).
Citation: Comparative Medicine 74, 5; 10.30802/AALAS-CM-24-033
Arterial pH.
A total of 169 studies were included in arterial pH analysis. The mean obtained by random effect models was 7.35 (95% CI: 7.34 to 7.35). Heterogeneity was significantly high at an I 2 of 99.6%. Subgroup analysis taking into consideration blood source, preductal (7.35; 95% CI: 7.35 to 7.36), postductal (7.35; 95% CI: 7.34 to 7.36), or missing (7.36 95% CI 7.35; 7.36), did not find any significant differences (P = 0.082). Similarly, gestational age did not have a significant effect. The funnel plot (Figure 6) showed a cluster of responses at the top of the funnel, with mean estimates between 7.3 and 7.4 for most studies. The pattern of responses was missing at the bottom of the funnel, suggesting that estimates about the means have a narrow range of values.
Citation: Comparative Medicine 74, 5; 10.30802/AALAS-CM-24-033
The Egger test for small study effects was found to be significant (P = 0.0002), supporting the idea of small study effects.
Umbilical blood flow.
The publications describing umbilical blood flow are depicted in Tables S1 to S3. A wide array of methods was used to measure blood flow including Doppler, implantation of flow probes, diffusion techniques, and cannulation of vessels (Table S2 includes those describing the values in mL/kg of fetal weight/min, and Table S3 includes publications that also describe the values in mL/min). We focused the analysis on values depicted in relationship to fetal weight.
Random effect analysis (Table S1) found a mean of 202.3 mL/kg/min (95% CI: 181.9 to 222.7), with heterogeneity quantification I 2 = 96.0% (95% CI: 94.9 to 96.9). The funnel plot (Figure 7) showed a number of studies falling outside the 95% CI with a broad range of error in some cases. The Egger test showed no significant small study effect (P = 0.75).
Citation: Comparative Medicine 74, 5; 10.30802/AALAS-CM-24-033
Venous blood gas values.
Only 6 publications reported venous blood gas values in fetal circulation, which are depicted in Table 2.
| Author and year | pH | pCO2 (mm Hg) | pO2 (mm Hg) | SatO2 (%) |
|---|---|---|---|---|
| Arroyo 2008: umbilical vein at 95 d gestation 3 | 7.26 ± 0.02 | 61.93 ± 2.06 | 48.63 ± 15.8 | 55.3 ± 0.061 |
| Arroyo 2008: umbilical vein at 130 d gestation 3 | 7.37 ± 0.01 | 45.9 ± 4.61 | 18.9 ± 1.47 | 52.2 ± 7.03 |
| Brown 2016: umbilical vein in late gestation 7 | 7.39 ± 0.01 | 44.5 ± 1.3 | 33.9 ± 1.9 | 81.7 ± 3.4 |
| Jones 2019: umbilical vein at approximately 125 d gestation 18 | 7.4 ± 0.01 | 43.58 ± 1.25 | NA | 79.14 ± 3.01 |
| Pena 2007: cerebral sagittal sinus near term 24 | NA | NA | 16 ± 1 | 43 ± 2 |
| Tomimatsu 2006: cerebral sagittal sinus at 122 ± 3 d gestation 30 | 7.33 ± 0.01 | 53 ± 1 | 16 ± 1 | 41 ± 1 |
Values are mean ± SD.
Blood volume.
Studies were screened for any description of fetal blood volume. A total of 5 studies had reported values. Two groups 16,32 described total fetal blood volume. Three studies reported fetal-placental volume. 4,19,21 Findings are described in Table 3.
| Author and year | Gestational age (days) | Blood volume |
|---|---|---|
| Jensen 2003 16 | 128–132 | 166 ± 17 mL/kg |
| Wolf 2001 32 | 124 ± 1 | 331 ± 16 mL |
| Wolf 2001 32 | 130 ± 2 | 342 ± 55 mL |
| Kusaka 2002 19 | 113 | 225 ± 31 mL |
| Matsuda 2006 21 | 111 ± 1.3 | 179 ± 27 mL |
| Matsuda 2006 21 | 110 ± 1.3 | 197 ± 14 mL |
| Assad 2001 4 | 125–145 | 216 ± 29.3 mL |
Values are mean ± SD.
Body weight.
Some studies have created formulas for the calculation of fetal body weight throughout gestation. Two studies attempted to validate the relationship between gestational age and body weight. One study 10 assessed 74 fetuses and got the polynomial equation to best describe the growth curve: fetal weight = 0.00082 × gestational age2 − 0.10462 × gestational age + 3.6508, where fetal weight is in kilograms and gestational age is in days (r = 0.94).
The second study 20 pooled together the data from 3 previous studies and reported a growth curve for sheep from gestational age 106 until 45 d after birth: BW = B0 + B1 × Age_GD, where BW represents body weight in kg, Age_GD represents gestational age values in days or days since the beginning of gestation, B0 = −5.4003, and B1 = 0.064 (R 2 = 0.846).
A different group 8 set out to estimate birthweight in fetal sheep by ultrasound analysis. They suggested that abdominal circumference, associated with renal volume were the parameters that most strongly correlated with weight.
While another study 11 performed a meta-analysis to establish the relationship between ovine fetal body weight and body length (CRL) and showed that the power term for formula expressing the exponential relationship between BW and CRL in late-gestation normal growing ovine fetuses is 1.5.
Conclusions
This study provides our best estimates of relevant variables on normal hemodynamics, blood gases, and acid-base status in the sheep fetus using statistical analysis that accounts for the high degree of study variability present in the literature.
To identify potential causes of statistical heterogeneity, we conducted subgroup analyses based on gestational age, year of publication, and source of blood without any significant changes in our results. Based on this observation, we concluded that the main source of heterogeneity is likely the limited size of most included studies. 29
Initial findings from investigating 44 fetal lambs in utero in 1970 suggested that low to no variation in fetal blood gas values is expected throughout gestation. 26 This observation aligns with the absence of important shifts when we performed the subgroup analysis by gestational age, when the only significant differences were noted in arterial pCO2 in term fetuses. Interestingly, they also evaluated a possible correlation with fetal weight and did not find any significant changes.
One group 27 performed an in vivo validation of fetal vessel T2 MRI oximetry and compared it to normal fetal SatO2 both in sheep and humans. A saturation value of 54 ± 6% (mean ± SD) was obtained in lamb arterial blood by their technique, which they validated by comparing with previous studies. They found a significant correlation between some vessel pairs in human and sheep fetuses, further suggesting the importance of this model for future studies. Later, the same group studied umbilical blood flow variations in sheep fetuses with noninvasive techniques and found that, for the most part, it does not vary with gestational age.
Furthermore, attempts to determine fetal body weight throughout gestation were reviewed. It has been suggested that lamb’s body weight varies little in the second half of pregnancy, 10 while others correlated advanced gestational ages with higher fetal weights, particularly after 106 d of gestation. 20 However, there was only one instance of agreement between their results regarding expected weights during gestation indicating that further investigation is warranted.
We recognize that in some studies, medications for sedation, anesthesia, or even respiratory support may have been used in the pregnant ewe; however, we believe that this would not significantly impact the obtained fetal values, since our analysis obtained data from control groups, it is expected that they would maintain fetal physiology. Furthermore, broad inclusion may increase the external validity of the values proposed. We believe that wider adoption of guidelines for reporting animal research such as the ARRIVE guidelines 25 will improve systematic reviews and produce accurate estimation of physiologic variables.
Our study provides pooled estimates of physiologic variables that are frequently monitored in animal models (sheep fetus). We hope that these estimates provide researchers with the tools necessary to evaluate fetal diseases and test innovative treatments. We advocate for researchers to use studies like ours for comparative analysis in their research, rather than relying solely on individual studies from the literature. However, if a single study shares comparable characteristics with the ongoing research, and upon evaluation demonstrates adequate methodology, as well as accurate and reliable findings, then it may serve as a suitable point of comparison. Overall, this review contributes to our understanding of sheep fetal physiology and calls for further investigation to address the limitations of the literature available. Having accurate estimates of fetal physiologic variables is crucial for the advancement of fetal diagnosis and treatment.
Strategy listing all search terms used and how they were combined.
PRISMA flow diagram. A flow diagram showing the process of eliminating studies that did not meet inclusion criteria.
Funnel plot of oxygen saturation values. Each study included is represented as a dot, y-axis is a measure of precision (SE), and x-axis is the mean. The dotted lines forming a triangle represent the 95% CI within which studies are expected to fall. The lighter vertical line corresponds to no intervention effect.
Funnel plot of arterial pCO2 values. Each study included is represented as a dot, y-axis is a measure of precision (SE), and x-axis is the mean. The dotted lines forming a triangle represent the 95% CI within which studies are expected to fall. The lighter vertical line corresponds to no intervention effect.
Funnel plot of arterial pO2. Each study included is represented as a dot, y-axis is a measure of precision (SE), and x-axis is the mean. The dotted lines forming a triangle represent the 95% CI within which studies are expected to fall. The lighter vertical line corresponds to no intervention effect.
Funnel plot of arterial pH values. Each study included is represented as a dot, y-axis is a measure of precision (SE), and x-axis is the mean. The dotted lines forming a triangle represent the 95% CI within which studies are expected to fall. The lighter vertical line corresponds to no intervention effect.
Funnel plot umbilical blood flow. Each study included is represented as a dot, y-axis is a measure of precision (SE), and x-axis is the mean. The dotted lines forming a triangle represent the 95% CI within which studies are expected to fall. The lighter vertical line corresponds to no intervention effect.
Contributor Notes
This article contains supplemental materials online.