To analyse the economic and resource implications of using plasma soluble fms-like tyrosine kinase-1 s(Flt1) and placenta growth factor (PlGF) measurements in pre-eclampsia evaluation and management.
To analyse the economic and resource implications of using plasma soluble fms-like tyrosine kinase-1 s(Flt1) and placenta growth factor (PlGF) measurements in pre-eclampsia evaluation and management.
Retrospective cost analysis of our prospective cohort study.
Boston, Massachusetts (USA).
Women (n = 176) presenting to the hospital at <34 weeks of gestation for evaluation of possible pre-eclampsia during 2009–10. Cases without complete cost or outcome data (n = 9) and re-enrolments (n = 18) were excluded.
Modelled comparisons between the standard approach (combination of blood pressure, urinary protein excretion, alanine aminotransferase and platelet counts) and a novel approach (ratio of plasma sFlt1 and PlGF) using actual hospital data converted to 2012 US dollars in accordance with the Centers for Medicare and Medicaid Services.
Direct 2-week costs and resource use by groups having true or false positive and negative test results for adverse outcomes according to approach.
The improved specificity of the novel approach decreased the proportion of women falsely labelled as test-positive from 42.3% (34.4–50.2%) to 4.0% (0.85–7.15%) and increased the proportion correctly labelled as test-negative from 23.5% (16.7–30.3%) to 61.7% (53.9–69.5%). This could potentially reduce average per-patient costs by $1215. Substantial quantities of resources [47.2% (35.7–58.7%) of antenatal admissions and 72.5% (68.0–77.0%) of tests for fetal wellbeing] were unnecessarily used for women who were truly negative. A proportion of iatrogenic preterm deliveries among women with negative results was potentially avoidable representing further cost and resource savings.
Clinical use of the plasma sFlt1 and PlGF ratio improves risk stratification among women presenting for pre-eclampsia evaluation and has the potential to reduce costs and resource use.
After more than a century of investigation, pre-eclampsia remains a major public health problem affecting roughly 5% of pregnancies yet contributing to nearly 15% of direct maternal deaths and 25% of perinatal/neonatal deaths worldwide. A disproportionate amount of resources and healthcare expenditure is dedicated to its identification and management. In 1992, approximately $20 billion was spent in the USA to treat women with pre-eclampsia and their neonates.[2, 3] Simply establishing the diagnosis currently requires resource-intensive measures such as hospital admission with close observation of maternal blood pressure, proteinuria and laboratory trends. However, this approach fails to predict the most serious adverse outcomes [eclampsia and HELLP (haemolysis, elevated liver enzymes and low platelet count) syndrome] in a substantial percentage of women.
The altered placental release of the anti-angiogenic factor soluble fms-like tyrosine kinase-1 (sFlt1) and the pro-angiogenic factor placenta growth factor (PlGF) has been proven integral to the pathogenesis of pre-eclampsia. Measurement of these factors in the maternal circulation correlates well with disease onset and progression.[6-9] We have recently shown that the plasma sFlt1/PlGF ratio predicts adverse outcomes occurring within 2 weeks among women with suspected pre-eclampsia presenting at <34 weeks of gestation. This ratio alone performed better than the standard battery of clinical diagnostic measures including systolic blood pressure, proteinuria, uric acid, alanine aminotransferase, platelet count and creatinine. The growing body of evidence supporting this test's utility has led to its adoption into standard clinical practice within some European hospitals; however, little is known about the impact of its use on resource utilization. This knowledge is crucial in determining its impact on public policy and implementation into practice in the USA.
Using data from a prospective cohort of women presenting with possible pre-eclampsia at <34 weeks of gestation, we sought to determine the economic and resource implications of using sFlt1 and PlGF measurements as a means of stratifying risk and management strategies over the subsequent 2-week period.
From July 2009 to October 2010, all women with singleton pregnancies who presented to the obstetric triage unit at the Beth Israel Deaconess Medical Center in Boston, MA, for evaluation of possible pre-eclampsia were offered enrolment. Indications for evaluation included elevated blood pressure, proteinuria or any symptoms associated with pre-eclampsia such as headache, visual symptoms, right upper quadrant pain or oedema. Standard tests and evaluation measures were performed at the discretion of the woman's attending physician. Discarded plasma samples from triage were collected from the clinical laboratory, processed and stored at −80°C. Automated analysis was performed at the clinical laboratory of Charité Hospital (Berlin, Germany) with the commercially available assays on the Elecsys platform (Roche Diagnostics, Penzberg, Germany) as previously described. All assays were performed after attainment of clinical data and the assay operators were blinded to the clinical information of the participants. The treating physicians were unaware of the sFlt1 and PlGF values.
All clinical outcomes were gathered over the subsequent 2 weeks and independently reviewed and adjudicated by two staff members before the availability of assay results. Adverse maternal outcomes were defined as the presence of hypertension plus one of the following: elevated aspartate aminotransferase or alanine aminotransferase (≥80 units/l), thrombocytopenia (platelet count ≤100 × 109/l), disseminated intravascular coagulation, abruption (clinical or pathological), pulmonary oedema, cerebral haemorrhage, seizure (in a woman without an underlying seizure disorder), acute renal failure (creatinine >132.6 μmol/l) or maternal death. The adverse fetal/neonatal outcomes included delivery indicated for hypertensive complications of pregnancy as reported by the primary obstetrician, small-for-gestational-age birthweight (≤10th centile for gestational age), abnormal umbilical artery Doppler waveforms (absent or reverse flow), fetal death and neonatal death. Two study staff members blinded to assay results performed the adverse outcome determinations. Diagnoses and adverse outcomes were ascertained within 2 weeks of presentation. The Beth Israel Deaconess Medical Center institutional review board approved the study, and all women provided written informed consent.
Participant re-enrolment was permitted among the cohort; however, determination of costs and resource use was restricted to participants' initial study encounter. Direct antepartum and delivery-associated resources and costs were determined by review of the participants' medical records and hospital administrative data. Actual institutional charges from all triage evaluations, maternal radiological studies, fetal radiological evaluations, tests of fetal wellbeing (including all combinations of biophysical profiles, nonstress tests, obstetric ultrasounds and Doppler ultrasounds), laboratory evaluations, admissions, consultations, deliveries and miscellaneous resources (intravenous tubing, intravenous fluids, etc.) were tallied over the 2-week period following enrolment. The reason for the 2-week window is given by the clinical utility of the test as described in the original study. Study personnel blinded to the sFlt1, PlGF and pregnancy outcomes extracted the data. All costs and charges were converted into 2012 US dollars using the conversion factor for 2012 (34.03) and the National Payment Amount for Geographic Practice Cost Index according to the Centers for Medicare and Medicaid Services. The cost of the test was fixed at $101.14 for measurement of both sFlt1 and PlGF and applied to groups in modelling the novel approach's budget impact.
The study is a cost analysis with a budget impact model. Given that the data were collected in a prospective cohort and were not used clinically, we could not perform a traditional cost-effectiveness analysis. Comparisons in cost and resource use were made between the ‘standard evaluation approach’ (including the combination of blood pressure, proteinuria, elevated transaminases: alanine aminotransferase or aspartate aminotransferase ≥80 units/l, and low platelet count: ≤100 × 109/l) and the ‘novel approach’ (measurement of sFlt1 and PlGF). We used our previously defined sFlt1/PlGF ratio ≥85 as ‘test-positive’ in the prediction of adverse maternal and fetal outcomes. This value demonstrated the strongest test performance on receiver operator characteristic analysis within this cohort and among others.[11, 16] Comparisons were made between testing strategies for groups of women with both true and false positive or negative results. Costs for the novel approach included the $101.14 per-patient cost increase attributable to the addition of the plasma sFlt1 and PlGF. Two separate budget impact analyses were modelled: an ‘ideal’ scenario and a ‘worst-case’ scenario.
The ideal scenario modelled the budget impact of two variables: the redistribution of women into groups based on test performance of the sFlt1/PlGF ratio, and the proposed cost savings among the new ‘true negative’ group. Here, the assumed costs represented the average cost of the initial triage visit alone. We argue that subsequent testing and/or delivery of these women in the following 2 weeks was not indicated and represented potential cost savings had the sFlt1/PlGF ratio been used clinically. The average cost of the enrolment triage visit represented the median 2-week cost among women in the ‘true negative’ group with only one hospital evaluation and no antepartum admissions. The worst-case scenario model made no assumptions regarding changes to management decisions regarding resource use or cost. Its budget impact relied solely upon the redistribution of women into groups based on the test performance of the sFlt1/PlGF ratio. A sensitivity analysis was then performed whereby the sFlt1/PlGF ratio cut-off denoting test-positivity and the cost percentage were varied to assess the budget impact of the novel approach versus the standard approach.
The perspective undertaken for the analysis was the hospital/third-party payer given that the large burden of costs associated with this disease lay with the payer. Much of the data regarding indirect costs and patient resources were not available, preventing an analysis from a societal perspective. Due to a short time horizon of 2 weeks after initial presentation, discounting was not undertaken. Sensitivity analyses were performed varying all costs from 50 to 200% of the baseline, including the cost of the test, to account for the variation in payment structure within the USA.
In all, 176 women <34 weeks of gestation were encountered during the study period. Of these, three were eliminated because of missing financial data, six were eliminated for missing delivery data and 18 were participant re-enrolments. A total of 149 women provided complete cost and resource data for review. Within 2 weeks following the initial hospital triage visits, adverse outcomes occurred in 34.2% of participants (n = 51). The distributions of participants by clinical outcome and test results for the standard approach and novel approach are shown in Figure 1. Demographic characteristics were relatively similar among groups defined by the novel approach; however, comparison of the initial clinical triage parameters by group correlated with enhanced risk stratification (Table 1). Women in the true positive group demonstrated significantly increased systolic and diastolic blood pressures (P < 0.0001), proteinuria (P < 0.0001), and laboratory findings associated with severe disease (Table 1).
|Variable||True positive||False positive||True negative||False negative||P-value|
|Gestational age (weeks)||30 (27–33)||31 (29–33)||31 (28–33)||32 (29–34)||0.47|
|Age (years)||32 (26–36)||34 (32–38)||32 (28–35)||31 (28–32)||0.53|
|Body mass index (kg/m2)||29.9 (27.5–34.6)||27.8 (25.1–31.2)||33.9 (30.3–37.8)||36.8 (29.7–41.4)||0.002a|
|Nulliparous||22 (56.4)||6 (100.0)||48 (52.2)||7 (58.3)||0.15|
|Smoker (%)||4 (10.3)||0 (0.0)||8 (8.7)||0 (0.0)||0.60|
|White/Caucasian||24 (61.5)||5 (83.3)||50 (54.4)||6 (50.0)||0.24|
|Black/African American||4 (10.3)||0 (0.0)||20 (21.7)||5 (41.7)|
|Asian||4 (10.3)||1 (16.7)||5 (5.4)||0 (0.0)|
|Other||7 (18.0)||0 (0.0)||17 (18.5)||1 (8.3)|
|History of pre-eclampsia (%)||7 (18.0)||0 (0.0)||18 (19.6)||4 (33.3)||0.39|
|History of chronic hypertension (%)||9 (23.1)||1 (16.7)||30 (32.6)||6 (50.0)||0.28|
|History of diabetes (%)||1 (2.6)||0 (0.0)||11 (12.0)||3 (25.0)||0.09|
|Clinical parameters in triage|
|Highest systolic blood pressure (mmHg)||152 (144–164)||130 (124–147)||133 (123–143)||145 (139–148)||<0.0001a|
|Highest diastolic blood pressure (mmHg)||97 (88–104)||86 (80–90)||83 (73–91)||93 (79–99)||<0.0001a|
|Proteinuria (>1+) (%)||29 (74.4)||1 (16.7)||17 (18.5)||7 (58.3)||<0.0001a|
|Alanine aminotransferase (units/l)||25 (14–26)||17 (14–30)||16 (12–23)||17 (14–23)||0.01a|
|Creatinine (μmol/l)||61.9 (53.0–70.7)||53.0 (44.2–53.0)||44.2 (44.2–53.0)||53.0 (44.2–53.0)||0.0003a|
|Uric acid (μmol/l)||374.7 (291.5–422.3)||333.1 (249.8–356.9)||243.9 (202.2–291.5)||279.6 (237.9–333.1)||<0.0001a|
|Platelet count (×109/l)||209 (143–240)||255 (167–315)||260 (227–298)||263 (195–357)||0.0002a|
|S/P ratio||348.0 (183.5–641.0)||145.3 (125.0–210.3)||4.0 (1.8–10.8)||12.1 (3.7–39.3)||<0.0001a|
|Time to delivery from triage (weeks)||0.3 (0.0–0.4)||4.2 (3.4–4.7)||6.4 (4.7–9.4)||0.6 (0.1–0.7)||<0.0001a|
|Gestational age at delivery (weeks)||30 (27–32)||35 (34–36)||37 (36–39)||32 (29–34)||<0.0001a|
|Birthweight (g)||1275 (780–1575)||2210 (1795–2390)||3115 (2568–3490)||1725 (1278–2175)||<0.0001a|
The novel approach demonstrated improved specificity over the standard approach (93.9% versus 35.7%). This decreased the proportion of women falsely labelled as test-positive (4.0% versus 42.3%) and increased the proportion correctly labelled as test-negative (61.7% versus 23.5%). Women demonstrating an sFlt1/PlGF ratio <85 without incurring any adverse outcomes over 2 weeks were categorised as truly negative and represented 61.7% of the total cohort (n = 92)—an expansion of 38.2% over the standard approach (Figure 1). Costs and resource data reflected the clinical decision making of physicians following the standard evaluation approach. However, by applying the novel approach and re-categorising participants, we were able to evaluate potential savings.
The true negative group represented the population where cost and resource savings could be realised. Despite significantly fewer ‘high-risk’ clinical parameters, these women were managed over the entire 2-week study interval with similar resources to those used in the truly positive women (Table 2). This group contributed 47.2% of the antepartum admissions, 72.5% of the tests for fetal wellbeing (including all combinations of biophysical profiles, nonstress tests, obstetric ultrasounds and Doppler ultrasounds), 43.8% of maternal imaging studies (magnetic resonance imaging, magnetic resonance angiograms, chest X-rays), and 40.2% of all serum and urine laboratory tests (serum chemistries, complete blood counts, coagulation studies, urinalyses and urine dipstick analyses). Resources used during the initial work-up were inevitable, but we proposed that subsequent resources and costs over the 2 weeks following an sFlt1/PlGF ratio <85 were unnecessary.
|Resource use||True positive (n = 39)||False positive (n = 6)||True negative (n = 92)||False negative (n = 12)||P-value|
|Triage visits||1 (1–1)||1 (1–1)||1 (1–2)||1 (1–2)||0.001a|
|Caesarean section (%)||33 (84.6)||4 (66.7)||55 (59.8)||8 (66.7)||0.05|
|Antepartum admission (%)||25 (64.1)||4 (66.7)||34 (37.0)||9 (75.0)||0.005a|
|Duration of antepartum admission (days)||3 (3–5)||3 (2–9)||4 (3–11)||3 (2–5)||0.41|
|Outpatient visits (%)||0 (0.0)||3 (50.0)||45 (48.9)||3 (25.0)||<0.0001a|
|Tests of fetal wellbeing||2 (1–3)||2 (0–2)||2 (1–4)||4 (2–5)||0.03a|
|Magnetic resonance imaging (%)||1 (2.6)||0 (0.0)||6 (6.5)||1 (8.3)||0.71|
|Magnetic resonance angiogram (%)||2 (5.1)||0 (0.0)||6 (6.5)||0 (0.0)||0.74|
|Chest X-ray (%)||8 (20.5)||0 (0.0)||2 (2.2)||1 (8.3)||0.003a|
|Complete blood count||8 (5–11)||2 (1–3)||1 (1–3)||7 (3–9)||<0.0001a|
|Coagulation panel||8 (4–14)||1 (1–2)||2 (2–4)||5 (2–8)||<0.0001a|
|Serum chemistries||37 (22–66)||7 (6–9)||8 (5–20)||24 (16–26)||<0.0001a|
|Urine studies||1 (1–2)||2 (1–2)||2 (1–4)||3 (1–4)||0.001a|
|Consultations||1 (1–2)||0 (0–0)||0 (0–0)||1 (1–2)||<0.0001a|
The median 2-week total cost for women in the true negative group who were evaluated only once in triage and were not admitted to the hospital was $370 (n = 46). Applying the cost of the novel test ($101.14 added to all participants' costs), the result ($471) represents the ideal cost for women in the true negative group, and we employed this in our model's ideal budget impact analysis (Table 3. The improved specificity of the novel approach redistributed the participants and average cost-per-patient tallies of the groups. By more precisely defining the true positive group, this shift produced a cost-per-patient increase ($4325 versus $3933). However, this increase was offset by cost savings from a reduction in the number and costs associated to false-positive participants (Figure 2). Most importantly, the implementation of the ideal cost upon the expanded true negative group contributed to net savings of $1215 per patient and $180 969 for the entire cohort.
|Cost resource||Standard approach||Novel approach||Budget impact|
|N||Total $||Cost per patient||N||Total $||Cost per patient||Total cohort||Per patient|
|True Positive||48||$188 787||$3 933||39||$168 660||$4325||−$20 127||$392|
|False Positive||63||$199 324||$3164||6||$17 988||$2998||−$181 336||−$166|
|True Negative||35||$48 137||$1375||92||$43 345||$471||−$4805||−$904|
|False Negative||3||$13 992||$4664||12||$39 277||$3273||$25 285||−$1391|
|Total||149||$450 239||$3022||149||$269 269||$1807||−$180 969||−$1215|
|True Positive||48||$188 787||$3933||39||$157 326||$4034||−$31 461||$101|
|False Positive||63||$199 324||$3164||6||$19 590||$3265||−$179 734||$101|
|True Negative||35||$48 137||$1375||92||$135 792||$1476||$87 655||$101|
|False Negative||3||$13 992||$4664||12||$57 180||$4765||$43 188||$101|
|Total||149||$450 239||$3022||149||$369 888||$2482||−$80 351||−$540|
Although the analysis of the ideal budget impact used real cost data, it relied upon an assumption that physicians would manage women with an sFlt1/PlGF ratio <85 without additional costs or resources beyond the initial triage visit. This represented a hypothetical cost-per-patient reduction of $904 for the true negative group, and it greatly contributed to the reductions in total cohort and average per-patient costs (of note, the actual costs of standard management and delivery were applied to women in the false negative group within the model). Relinquishing this assumption, we calculated a ‘worst-case’ scenario in which per-patient costs for each group in the novel approach matched the corresponding values from the standard approach plus the $101.14 cost for the sFlt1/PlGF ratio (Table 3). This scenario assessed the budget impact of improved characterisation of participants without assumed changes in management strategy beyond the acceptance of the new risk characterisation. Total cohort costs were reduced by $80 351 and the cost-per-patient was reduced by $540.
Sensitivity analysis was performed for cohort and per-patient costs in the novel versus standard approaches by varying the cut-off sFlt1/PlGF ratio and retrieved percentage of cost. As demonstrated previously, the ratio of 85 demonstrated optimal balance of sensitivity and specificity (see Supporting Information, Table S1).[10, 11, 16] Net savings were predicted for both total cohort and per-patient costs with application of the novel approach using ratios between 5 and 200. In addition, model costs were varied between 50 and 200%. Cost savings were predicted with clinical application of the novel approach using all modelled cut-offs or cost percentages (Table 4).
|Cost resource||Cost ($) at 50%||Cost ($) at 75%||Cost ($) at 150%||Cost ($) at 200%|
|True positive||−12 036||−18 054||−36 108||−48 144|
|False positive||−90 972||−136 457||−272 915||−363 887|
|True negative||−7055||−10 582||−21 165||−28 220|
|False negative||12 036||18 053||36 107||48 143|
|Total||−82 957||−131 970||−279 011||−377 038|
Twelve women experienced adverse outcomes despite an sFlt1/PlGF ratio <85, and this group (false negatives) was predominantly obese with one half carrying a diagnosis of chronic hypertension (Table 1). Review of the clinical characteristics and outcomes of these women (see Supporting Information, Table S2) highlights the complex medical history and complaints that prompted preterm delivery in all of these women. The only adverse outcome was preterm delivery in ten of the twelve women, and subjective concerns such as headache and physician or patient desire was cited as the delivery indication in seven of the ten. Among the other three women, two were delivered due to nonreassuring fetal status with chronic hypertension and the third woman was delivered secondary to acute fatty liver of pregnancy and elevated liver function tests (a separate disease entity from pre-eclampsia). The two remaining adverse outcomes were one neonatal death in a woman after chronic placental abruption and prolonged oligohydramnios from 16 weeks of gestation and one woman with acute renal failure in the setting of chronic hypertension and diabetic nephropathy. As actual antenatal and delivery-associated data were used in the analysis of this group, no adjustment or speculation of the costs and resource use was performed to account for the potential implications of a ‘missed diagnosis’.
Neonatal outcomes among live-born infants experiencing iatrogenic preterm delivery within 2 weeks of study enrolment further demonstrate the sFlt1/PlGF ratio's ability to appropriately characterise risk (Table 5). Of the iatrogenic deliveries indicated for hypertensive complications of pregnancy, 76.6% occurred among women with an sFlt1/PlGF ratio >85. The remaining 12 occurred among women in the false negative group, as described above. All required admission to the neonatal intensive care unit, and the median length of stay was 17 days. Rates of additional morbidities were similar to neonates delivered to women from the true positive group (Table 5).
|Neonatal outcomes||True positive (n = 36)||False positive (n = 0)||True negative (n = 0)||False negative (n = 12)|
|NICU admission (%)||36 (92.3)||0 (0.0)||0 (0.0)||11 (91.7)|
|NICU (days)||18 (11–56)||0 (0.0)||0 (0.0)||17 (10–22)|
|Respiratory distress syndrome (%)||22 (56.4)||0 (0.0)||0 (0.0)||6 (50.0)|
|Intraventricular haemorrhage (%)||3 (7.7)||0 (0.0)||0 (0.0)||0 (0.0)|
|Necrotising enterocolitis (%)||1 (2.6)||0 (0.0)||0 (0.0)||0 (0.0)|
|Sepsis (%)||1 (2.6)||0 (0.0)||0 (0.0)||0 (0.0)|
|O2 requirement (%)||27 (69.2)||0 (0.0)||0 (0.0)||10 (83.3)|
|Perinatal mortality (%)||0 (0.0)||0 (0.0)||0 (0.0)||1 (8.3)|
In this retrospective cost analysis, we demonstrated potential reductions in direct hospital costs and resource use with implementation of the plasma sFlt1/PlGF ratio for pre-eclampsia evaluation and management. Although current standard diagnostic tests are readily available and engrained into routine practice, they lead to significant misclassification and over-testing/treating of many women. Large quantities of resources and costs are erroneously allocated to many low-risk women. Our results reveal that the increased specificity of this novel approach more accurately defines the population at risk and could enable appropriate and reduced cost/resource expenditure. Ideal use of this novel approach potentially yields large per-patient cost savings ($1215); and even the ‘worst-case’ scenario (or reduced percentages of the costs included in this model) could yield substantial per-patient savings ($540).
The characteristics of this test highlight its potential to serve a diagnostic purpose rather than as a screening tool. Women in this study represented a ‘pre-selected’ population in whom concerns for pre-eclampsia included varying degrees of hypertension or proteinuria, which prompted hospital evaluation and an invitation to participate. This ‘screening’ process is the basis of modern prenatal care and appears to work—demonstrating fair sensitivity in published reports and among our cohort. Unfortunately, women with varied post-test odds for true adverse outcomes are undergoing similar management following initial identification, and our results among true-negative participants demonstrated this. We envisage that clinicians will use the sFlt1/PlGF ratio in the triage setting on women who have already ‘screened-positive’ by standard antenatal care criteria. A ratio <85 in the absence of other overt clinical concerns may identify a candidate for expectant management via routine prenatal care. Those with ratios >85 represent women at significant risk for adverse outcomes and warrant the increased costs, resources and potentially indicated delivery associated with heightened surveillance.
Despite the potential major impact upon public health and healthcare expenditure, few studies have addressed the economics of pre-eclampsia evaluation and management. Our report uses actual costs and resources in comparing different approaches to pre-eclampsia evaluation and management. Hadker et al. developed a decision-analysis model that similarly assessed the economics of the addition of serum sFlt1 and PlGF measurement into standard obstetric practice of the UK NHS. The authors demonstrated that the novel approach provided substantial cost savings for the NHS, but their model was built upon theoretical data and assumed test characteristics from other publications. They applied the sFlt1/PlGF measurement at 20 weeks to all women, an approach that has been shown to demonstrate poor predictive value.[18, 19] Our application of the test to women with clinical signs and symptoms of pre-eclampsia parallels the approach by other studies where measurement of angiogenic factors has been shown to correlate strongly with diagnosis and outcome.[9, 10, 20] Hence, we propose that the assumed cost-savings demonstrated in the study by Hadker et al. could be magnified with a more appropriate application of this test to an at-risk population that undoubtedly has greater disease prevalence.
Strengths of this study include its use of actual direct hospital costs, clinical parameters and outcomes gathered in a prospective fashion. This allowed for calculation of true test characteristics and implementation of these values into cost and resource analyses. Limitations of this work include its small sample size and the generalisability of its findings. Our data come from a single tertiary care hospital in the USA representing the clinical practices of a single group of physicians. All costs are given in US dollars. Whether the same cost and resource savings could be achieved in other settings is purely speculative.
The observational nature of this study and the blinding of treating physicians to sFlt1/PlGF ratios limited our ability to fully evaluate cost and resource implications without making assumptions. As such, we attempted to create ‘ideal’ and ‘worst-case’ scenario models to account for the lack of actual cost/resource data from true clinical use of the novel approach. However, we cannot fully account for the real-time clinical application of this test specifically with regard to using the cut-off of 85 and the false negative diagnoses associated with this cut-off. Although our data suggest that indicated preterm delivery based solely on a diagnosis of pre-eclampsia in the absence of an elevated sFlt1/PlGF ratio may not be warranted, we cannot account for the various additional rationales for preterm delivery. Additionally, implications of a ‘missed diagnosis’ from the test's lack of 100% sensitivity could not be assessed within the cohort undergoing standard management.
Another limitation was the retrospective nature of data collection. Assessment of indirect costs such as out-of-hospital costs, financial implications on participants' activities of daily living, loss of income, childcare costs, long-term implications on maternal and child health, and the social burdens associated with hospital admission are best attained with prospective data collection tools employed by larger prospective studies.
Measurement of the serum sFlt1/PlGF ratio improves risk stratification among women presenting for pre-eclampsia evaluation, and its clinical use has the potential to reduce hospital costs and resource use among women at low risk for adverse outcomes. Whether the real-time use of the sFlt1/PlGF ratio in the triage and management of pre-eclampsia will result in reduction of iatrogenic preterm deliveries without increased adverse maternal, fetal or neonatal outcomes (with consequent direct and indirect cost/resource implications) needs to be proven. Performance of larger prospective cohort or randomised studies using these angiogenic factors as clinical determinants of risk and management strategy in real time is warranted.
WTS, DD, SJR and SR contributed to conception and design of the study; WTS, SS and SR contributed to acquisition of data; WTS, DD and JW contributed to data analysis and WTS, DD, JW and SR to interpretation of data. The article was drafted by WTS and SR and WTS, DD, JW, SS, SJR and SR revised it critically for important intellectual content.
The study was approved by the Beth Israel Deaconess Medical Center Institutional Review Board, and all women provided written informed consent.
Funding was provided by the Harvard Diversity and Community Partnership Faculty Fellowship and NIH/NICHD.
SR is supported by Harvard Diversity and Community Partnership Faculty Fellowship Award and K08HD068398-01A1 (NIH/NICHD). We thank Dawn McCullough, RN for recruitment of women and for data collection.