When it comes to the assessment and development of novel diagnostic tests, such as a test to identify a potential nonviable pregnancy, the key tenets are validation, validation, validation. In the article entitled, “Sensitivity and specificity of a urine screening test for first trimester abnormal pregnancy in an emergency setting,” Texiceria et al had the resolve to attempt to validate and to publish a study to demonstrate the lack of utility of a commercially available novel diagnostic test. There are 2 messages here. The first is that particular urinary-based tests that assess ratios of intact human chorionic gonadotropin (hCG), β core fragments, nicked β fragments, and β-hCG isoforms do not work particularly well in a busy emergency room setting. Specifically, the authors noted only 13% sensitivity, 36% positive predictive value, and 47% accuracy. Of note, this study of 800 women was larger than the study of 254 pregnant women that was used initially to demonstrate 97% sensitivity for ectopic pregnancy (EP).

The second message is the important realization that many purported diagnostic tests simply do not work as well in real world practice as they appear to work in the first study that established their potential utility. When developing a novel diagnostic test, and in particular the case of a nonviable first-trimester pregnancy, validation is paramount. Of importance is validation in a population that is distinct from the population in which the test was developed. This is not the first diagnostic aid for women at risk for an EP that has not validated well in an independent population. It likely will also not be the last. However, that does not mean we should stop trying.

Teixeira et al chose to evaluate independently a diagnostic test in a particular clinical conundrum of identification of an early abnormal gestation. Although the current clinical standards to evaluate a woman at risk for an EP or miscarriage has evolved over the past decade, we continue to have clinically unacceptable misclassification rates. In fact, “we” as busy clinicians are perhaps relying too heavily on diagnostic rules of thumb. Although understanding the expected rise (or fall) in serial hCG levels has allowed us to streamline the identification of women at risk for EP and miscarriage, these evaluations are not diagnostic. In fact, strict adherence to these rules can result in iatrogenic intervention and potential interruption of a desired intrauterine pregnancy (IUP) and the chance of missing an EP that is growing “normally.”

Up to one-half of all pregnant women have signs and symptoms, such as pain or bleeding, that are suggestive of a pregnancy loss in the first trimester. In addition, many asymptomatic pregnant patients carry nonviable gestations, some of which are EPs. When the gestation is not visible with ultrasound scan, distinguishing a healthy ongoing IUP from a miscarriage or an EP poses a critical clinical challenge because there is no other definitive noninvasive diagnostic test. Because of the particular concern that women with an EP are at risk for rupture and life-threatening intraabdominal hemorrhage, clinicians must monitor patients closely over several days to weeks until the location (and viability) of the pregnancy becomes apparent. Prompt diagnosis would permit earlier treatment and, thus, allow for procedures that preserve fallopian tube function and fertility and minimize the need for intervention for a miscarriage. The development of noninvasive methods of diagnosis to differentiate accurately an ongoing viable IUP from an EP or early pregnancy failure would improve clinical treatment of these patients and decrease morbidity and cost. The fact that the test under study does not appear to meet these criteria should not dissuade us from continuing the search for new tests for this important clinical conundrum.

Making a final diagnosis with a noninvasive test based on a biomarker is a clinical challenge. Ultimately, a woman will be diagnosed with 1 of 3 outcomes: a growing IUP, a spontaneous miscarriage, or EP. Currently, making a definitive diagnosis in a woman with a pregnancy of unknown location is cumbersome and requires multiple office visits, serial hCG measurements for up to 6 weeks, multiple ultrasound examinations, and, at times, surgical procedures (such as uterine curettage and/or laparoscopy). Most large clinical practices utilize a specialized electronic medical record (often called a “Beta book” or a “Quant book”) to observe women at risk until definitive diagnosis or resolution of the pregnancy occurs. This approach results in a high clinical volume. At our institution 4-8 women each day are determined to be at risk for EP, with up to 40 women under surveillance at any given time. Treatment for each outcome is drastically different, which makes the accuracy of diagnosis imperative.

The use of biomarkers in medicine is increasing. A molecular biomarker is a molecule that is produced by an affected individual that signals a specific clinical condition or disease state. The central laboratory business model is changing rapidly, with clear focus toward individualized and specialized sample analysis (personalized medicine) as part of point of care diagnostics. As clinicians, we will be inundated with the introduction of customized assays for specialty clinical laboratories in our hospitals, emergency departments, and our own medical offices. Biomarkers are being developed for many potential uses that include better diagnosis to identify differences in susceptibility and/or to enhance prediction of outcome.

Biomarkers widely are used successfully in a number of medical disciplines. For example, in cardiology, cardiac tropinin I or T and/or myoglobin are used to assist in the diagnosis of heart attack or to identify cardiac tissue damage. Biomarkers are being explored for use in kidney disease, thrombocytopenia, Alzheimer’s disease, prediction of pneumonia, and diagnosis and/or prognostic of various cancers.

The use of biomarkers in reproductive medicine is increasing and has included identification of a single nucleotide polymorphism in CYP3A4, an enzyme in the cytochrome P450 system, and its association with risk of ovarian failure in breast cancer patients who undergo chemotherapy. Rebbeck et al has shown that steroid hormone metabolism genotypes predict menopausal symptoms from hot flashes to depression. Biomarkers that are used in reproductive biology often focus on diagnosis. Markers that are associated with different pathways, used in combination, have been a successful strategy. Perhaps the most common example is the quad screen used to assess the risk of a chromosomal anomaly in a fetus. Risk is based on levels of 4 markers (alpha-feto-protein, hCG, estriol, inhibin A). Recently OVA1 has been approved by the Food and Drug Administration and is available for clinical practice. This test utilizes 5 biomarkers (transthyretin, apolipoprotein A-1, beta 2-microglobulin, transferrin, and cancer antigen 125) to determine the likelihood of malignancy in women with an ovarian mass for whom surgery is planned. A multivariate biomarker for noninvasive diagnosis of endometriosis is on the verge of clinical use.

The development of a biomarker panel to aid in diagnosis of women at risk for EP must fallow a rigorous process. Extensive testing, validation, and modification must be performed before a biomarker is demonstrated to have clinical utility. If a marker is near, or is ready, for clinical practice, its intended use, strengths, and limitations should be elucidated clearly and understood. The transition of a biomarker to clinical utility is an exacting and iterative process. There are no short cuts. There is much to be learned about the predictive significance of proposed markers before any of these goals can be realized. We must be very wary of the “1 and done” publications of a new marker that never proceed to validation.

We should not stop looking for new tests for women at risk for a nonviable gestation or EP because a new urine test does not validate its marketed utility. The use of biomarkers for diagnosis of EP can change the diagnostic and therapeutic paradigm. A biosignature for EP and/or nonviable early gestation could be used in multiple ways. If differential expression of proteins identifies an extrauterine pregnancy, clinical care would be changed immediately that would allow for early noninvasive medical intervention. Also of importance would be a biosignature profile that identifies a nonviable gestation (including miscarriage and ectopic pregnancy). If the possibility of ongoing IUP is eliminated, a clinical strategy could be developed to allow early medical intervention, irrespective of location. If an ongoing viable IUP is predicted accurately, a woman can be safely triaged for obstetric care, and the interruption of a desired pregnancy in the course of diagnosis or treatment of EP can be minimized, if not eliminated. This important diagnostic error and the inappropriate use of the teratogen and abortifacient methotrexate have been noted repeatedly and independently by investigators as an iatrogenic complication of contemporary treatment of women at risk for EP.

Finally, it is possible that a biosignature could distinguish an aggressively growing, but nonviable, gestation (even if the location is not known with certainty) that would allow a clinician to better determine whether optimal management is expectant, medical, or surgical. The specific clinical strategies will depend on the results obtained. However, one easily can envision that the care of women with all outcomes of a symptomatic early pregnancy can be improved. Either of these strategies could decrease dramatically the frequency of outpatient surveillance and the use of invasive and costly diagnostic tests and procedures.

The current state of the science regarding biomarkers for EP is disparate. Potential biomarkers are in various stages of development. Many serum markers have been proposed to aid in the identification of an EP, but few have been validated. Moreover, few have been evaluated in combination. Paramount to the transition into clinical practice is validation in populations that include all potential outcomes for patients with a symptomatic first-trimester pregnancy and, thus, must also include miscarriage.

The identification and development of a biomarker has well-characterized and distinct phases. Phase I is preclinical exploration to identify promising markers. There are 2 general paths in the identification of a biomarker that include (1) assessment of likely markers based on our understanding of the mechanism (in this case, understanding normal and abnormal implantation) and (2) unbiased candidate biomarker discovery. Phase II is establishing a clinical assay to be used on a larger scale. Phase III is testing the utility of the biomarker, often by retrospective case control studies. Phase IV is validation of the biomarker with the use of prospective screening to quantify the clinical value of the biomarker test in the population of its intended use.

Again, I would like to compliment the authors for conducting and publishing a high-quality phase IV validation study. Although it did not demonstrate validation in their population, a well-done study, even with negative findings, deserves to see the light of day. These data will inform both the progression of the science and will improve the care of our patients.