Zika, Exploration, and the Null Hypothesis

On the Practice of Ultrasound

Navel Strategy
This article has been prompted by a spectacular failure of ultrasound to provide substantive information about the intracerebral consequences of Zika infections during pregnancy in initial, peer-reviewed reports.

There are some insights to be gleaned, and I will show you some obvious, 30-year-old techniques that will improve your ability to evaluate fetal brains by a zillion percent. This will be a belly button analysis. The fetal brain is the “outie;” the “innie is molded by the inertial restraints that impact all clinical applications.

The Great Ultrasound Divide
Ultrasound imaging practice can be subdivided into the exploratory camp that asks “what do I need to see?” This was common in the early days of the field; it was guided by the gross pathology and pathophysiology of diseases that we sought to depict with our new equipment. The other and vastly more common schema operates on the basis of “what is seen” during an exam. It tends to be limited to gross anatomy and follows from a rigid protocol, like forming transverse and longitudinal scan planes through some part of the body.

This form of ultrasound has enabled spread of the technique with minimal education requirements for its users, albeit at the cost of retarding innovation and extending the clinical utility of what we do.

Zika
Zika has emerged as a major reproductive health crisis, relentlessly sweeping through South America, Central America, and the Caribbean, and soon to seek asylum in North America. The mosquitoes that vector the virus between people can survive year-round from Texas through South Carolina and in summer months along the east coast, north to New York. The structure of the Zika virus has just been reported in Nature as an advanced online publication.

We all know that it took extended time and a massive increase in incidence of microcephaly for the link to be exposed in Brazil and shared with health agencies. In retrospect, there had been an increased incidence in microcephaly in Micronesia, the intermediate stop in the pathway of Zika from Africa to Brazil, less evident from available health statistics there. Virulent strains with progressively increased attack rates in pregnancy have developed as the virus has migrated into a substantially different population.

Microcephaly is a visible, gross finding with a numerical marker from using a tape measure. Galen’s tailor must have used a tape measure for his togas 2000 years ago, and it was not a new invention then. It does seem ironic that so many diagnostic things we do every day in our era of instant genomics and molecular marvels still rely on simple measurements.

Microcephaly and Zika are now linked causally. Zika passes through the placenta. A basic understanding of fetal brain development implies that microcephaly results from infections in the late first or early second trimester. The obvious questions are what lesser, but still devastating effects on cerebral development and function may occur from infections of pregnant women after 20 weeks or in newborns? Is Zika neurotoxicity limited to the brain or parts of the brain (and eyes) or does it affect any or all of the rest of the developing nervous system or other organs (like the placenta)?

I was hoping for some basic clarification of these questions when the NEJM announced the expedited on-line publication from FIOCRUZ in Rio March 4 of this year. There was a case report from Slovenia published in the Journal, March 10. All featured ultrasound, none achieved satisfactory visualization of the brain, nor did they assess any behavioral clues, such as normal or pathologic motoric patterns, eye movements, auditory responsiveness, or habituation. My recollection is that ultrasound imaging of fetal eye movement patterns was reported in Science in 1978, hardly cutting edge now.

There have been some additional early reports with ultrasound that I will not review separately. They all seem to suffer from the same basic limitations in ultrasound technique. One of these was from my old alma mater, a case report of a woman infected at 11 weeks who had normal ultrasounds at 13, 16, and 17 weeks and an absolutely abnormal exam at 19 weeks. A 20-week MRI provided a lot more information, including a corpus callosal length equivalent to a 13-week growth arrest. There was one power Doppler image with the wall filters set so high that the only vessel that lit up was the portion of the pericallosal artery oriented along the beam axis. When I saw this, I thought that it was a double pity that not only was color Doppler sensitivity way off the mark, but even if it had been set up properly, it would not have been as descriptive as supersensitive color Doppler from plane wave insonification (Supersonic Imagine, introduced at the last RSNA). The virology and molecular biology work for this article were completely first rate.

The Rio Study
I really wish that the primary investigators at FIOCRUZ had opted for MRI for the women they had enrolled in their study. The ultrasound was done by perinatologists identified as certified by a specialty college and a society, which is unprecedented for a major research publication and uninformative.

These worthies used a GE Voluson E6 unit (performance dependent on arbitrary presets) limited to transabdominal imaging with a 4-8 MHz transducer. There is a long appendix to the article in which are listed all of the basic things one did in survey exams many years ago.

All of the exams were done at the examiners’ convenience in Rio. Twenty-eight of the original 70 enrolled women could not, or would not, travel to be examined, a study failure rate of 40%. The equipment is mobile. Why not bring the unit to the province where the patients resided? For that matter, for a groundbreaking study, as this ought to have been, the examiners could have requested loan of high performance equipment and consultants from any manufacturer. Even better, they could have demurred and permitted someone else in the country with practical experience with fetal or neonatal neurosonography to participate in the investigation.

There was one case of microcephaly, confirmed by CT after delivery, and there were four instances of calcifications in women with later pregnancy infections. There did not seem to be any kind of newborn or follow-up neurobehavioral exams in those identified as “normal” by ultrasound.

Contributions from Ljubljana
Each of the reports had a single ultrasound image, which was a base plane view. This avoids imaging pretty much all of the brain except for the brain stem, some cisterns, the Circle of Willis, and part of the cerebellum and leptomeningeal space. Ventricles will appear when they are massively dilated. In addition, there is signal attenuation by bone, distortion from refraction, and, with noisy systems, ring-down that precludes visualizing much of anything on the side near the transducer.

The Slovenian case report is of a woman who contracted Zika at the start of the second trimester in Brazil and returned home later in the pregnancy. There were 14 and 20 week exams in Brazil and one after her return home at 29 weeks, all said to be ”normal.” The patient noted decreased fetal movements, prompting a referral to the Department of Perinatology. The expert exam was performed three weeks later, identifying growth retardation, head circumference below the second percentile (solid tape measure work there) placental and intracerebral calcifications, mild ventriculomegaly, and a small transcerebellar diameter. And there was a sentence that I find inexcusable: Brain structures were blurred, and there were numerous calcifications in various parts of the brain.

The case report was invaluable for its main subject, the postmortem neuropathology.

The representative brain section photographs are coronal sections, there was agyria, and calcifications were identified as clusters of destroyed neurons. There were no other significant gross autopsy findings outside the brain and spinal cord.

The Fetal Brain
You never want to assume anything much about the brain from just the size of the cranium, although extreme variations are apt to be seriously problematic. The power of ultrasound is that you can have excellent visualization of the brain itself all throughout pregnancy and continuous imaging provides correlative functional information.

Many examiners use a base plane to get the head circumference measurement. It is not clear to me how this practice started, because it really does not make sense. The fetal skull base has a reasonably elliptical shape. This is characterized fully by long and short diameters. Fitting an ellipse to base planes that are not very carefully aligned introduces an error related to the area of the mismatch, which is like squaring the linear error you get from a diameter estimate.

I suppose that one learns to do a head circumference and keeps doing it, never questioning why this is being done and never posing the all-important query: is there a better way to get at what you really want to know? Long bone lengths are always better for gestational age determination. And, if one wanted a measure of cranial volume, then the distances from the skull base to the top of the skull needs to be included. Fixating on base planes precludes use of coronal and midline sagittal planes, which are excellent for BPD and OFD and provide a lot more info.

Once you start thinking in those terms, then it is a teeny tiny step to look within the skull at the brain. After the second trimester, positioning the probe right over the fontanelle will minimize distortion. Imaging with a high frequency, broad bandwidth endovaginal probe will have the lowest noise and the best spatial and contrast resolutions, and this portal always takes precedence over multilayered transabdominal viewing, whenever the embryo or fetus is positioned accordingly. The images below are really old, mainly from the cache from a defunct website I did in the 1990s, which I have been able to retrieve while on vacation.

[[{"type":"media","view_mode":"media_crop","fid":"48580","attributes":{"alt":"Coronal view","class":"media-image media-image-right","id":"media_crop_9670676846641","media_crop_h":"0","media_crop_image_style":"-1","media_crop_instance":"5807","media_crop_rotate":"0","media_crop_scale_h":"0","media_crop_scale_w":"0","media_crop_w":"0","media_crop_x":"0","media_crop_y":"0","style":"height: 358px; width: 625px; border-width: 0px; border-style: solid; margin: 1px; float: right;","title":"Figure 1. Coronal views early and later second trimester. Look at the obvious differences in gray scale features when neurons are proliferating from germinal periventricular zones (nourished by sugar in CSF and later after migration along radial glia and intracerebral cerebral circulation is well established).","typeof":"foaf:Image"}}]]

Calcifications, and Speculation, and the Corpus Callosum
Brain injury with Zikais a sequence of infection and inflammation, leptomenigitis followed by encephalitis. It seems that only some fetuses are compromised after maternal Zika infection, but this is a very muddy area. Early pregnancy infections will have maximal consequences, just like drug, radiation, and other exposure events.

Calcium deposition takes time; it occurs in clusters of destroyed neurons and residual Zika viral bodies. This finding is always significant clinically, but because of its timing, it is not an early screening marker. Calcifications are scattered, a geographic pattern rather than a global one. Think about skip areas of normal tissue in the gut with inflammatory disease, about focal areas of neoplastic transformation, or skin rashes with lesions separated by or surrounding normal appearing skin. What is the common factor that predisposes some cells to injury or infection, sparing seemingly identical adjacent cells?

Studies of the transcription architecture of the brain have identified separate genetic differences within fields of histologically indistinguishable neurons. Much of this work has come from the Allen Institute, as in Nature.

My main reason for bringing up this topic is a question of a subtle, educational one: can a physician tasked with evaluating a fetal brain be considered expert without an up-to-date familiarity with recent, substantive advances in the neurosciences?

[[{"type":"media","view_mode":"media_crop","fid":"48582","attributes":{"alt":"Midline sagittal magnification view","class":"media-image media-image-right","id":"media_crop_3600268592580","media_crop_h":"0","media_crop_image_style":"-1","media_crop_instance":"5808","media_crop_rotate":"0","media_crop_scale_h":"0","media_crop_scale_w":"0","media_crop_w":"0","media_crop_x":"0","media_crop_y":"0","style":"height: 508px; width: 625px; border-width: 0px; border-style: solid; margin: 1px; float: right;","title":"Figure 2. Midline sagittal magnification view of the corpus callosum. Old readers will recognize the antique XP marker.","typeof":"foaf:Image"}}]]

The corpus callosum is the main axon bundle connecting isologous regions of both hemispheres. It is the single, very best indicator of anatomic development of the cortex. The corpus callosum is very thin or absent with global compromises of cortical neuron proliferation and migration, and it is notched or locally thinned when there are focal injuries. In the mid-1990s, I reported normative length and area measurements for a few hundred ultrasound exams. Satisfactory visualization was achieved at least 90% of the time, and that was years ago with the equipment of the times. The anterior genu of the corpus callosum is visible from about 12 weeks, extending posteriorly with the rostral end fully formed by about 22 weeks GA.

One of the big trends in brain development in the third trimester is the great increase in surface area that occurs with sulcation. Sulcal pattern is another marker of anatomic development. Atlases of normal sulci appearance by gestation are available for correlation or one can seek disruptions in gyral architecture.

[[{"type":"media","view_mode":"media_crop","fid":"48583","attributes":{"alt":"Magnification midline sagittal view","class":"media-image media-image-right","id":"media_crop_5763446414653","media_crop_h":"0","media_crop_image_style":"-1","media_crop_instance":"5809","media_crop_rotate":"0","media_crop_scale_h":"0","media_crop_scale_w":"0","media_crop_w":"0","media_crop_x":"0","media_crop_y":"0","style":"height: 527px; width: 625px; border-width: 0px; border-style: solid; margin: 1px; float: right;","title":"Figure 3. Magnification midline sagittal view of sulci just above the corpus callosum.","typeof":"foaf:Image"}}]]

There was a landmark study of 15,000 plus antenatal ultrasound exams. The study design was comprehensive and the statistical analysis unimpeachable. The conclusion was that there was no outcome benefit to routine examinations over those with specific indications. This study has never been refuted; the response of the obstetrical community was to increase routine scanning. Of course, you might say that establishing ”normal” for a pregnant woman or an expectant couple is priceless. I believe that myself. The clinker is that in the RADIUS study, performance for detecting congenital anomalies was abysmal, so that normal was not all that definite, even for straightforward anatomic imaging.  Finding or excluding congenital heart disorders was a particular problem. The practical solution seems to have been to defer patients for fetal echocardiography provided by a pediatric service instead of requiring additional training for personnel doing more general fetal scanning.

My experience with examining a whole lot of fetuses is that evaluating the heart is really easy. Brains are much trickier, particularly since you do not derive additional functional information from Doppler, and because of the complexity of the brain below the resolution limits of ultrasound.

The initial Zika papers lead to the inescapable conclusion, enshrined in the very best of medical journals, that ultrasound is of no value for screening, nor is it good for anything less than the catastrophic occurrence of microcephaly late in pregnancy. I do not think that is true, although it may be so. What really bothers me is that the failure of ultrasound is attributed, as usual, to the method and not to the equipment or technique deficiencies of its users.

The Society of Maternal Fetal Medicine has just published a position statement on the subject. My heart sank when I saw the title: Ultrasound Screening for Fetal Microcephaly Following Zika Virus Exposure. The focus is on detecting end-stage terrible disease, which is of no practical value to a pregnant woman, who is terrified by potential Zika exposure. The statement does mention that neurosonography might be of value but does not elaborate further. The focus is the old sonic tape measure around the skull base. There is a reference to a paper in preparation from Israel that 40% of 42 cases identified as microcephaly in utero had normal head circumference at birth. There are only three possibilities, equipment calibration had never been checked, the measurements were unbelievably inept, or the reference standard was not appropriate.

The recommendations are also interesting, to wit, if head circumference is pretty low, but the anatomy is ”normal,” recheck in a few weeks, and if head circumference is way low, get neurosonography (which you should have been doing in the first place) three or four weeks later. I find this somewhere between disgraceful and sad. This is potential crisis situation with life-altering consequences, it is time to start thinking of the best we can do, not a lowest common denominator tactic that abandons a patient to diagnostic limbo.

One of the big problems is sorting out who has been infected (80% of exposures are asymptomatic) and of those in pregnant women, when are Zika virus particles in or through the placenta? A Harvard/MIT project for identifying RNA viruses from their genomes works for Zika. A color change reaction from sensors embedded in a paper disc has been proposed for rapid screening. What interests me is the potential for this and other procedures for rapid identification of Zika in amniotic fluid. We and others offered amniocentesis at risk levels <1:1000 in the era before cell free DNA.

The Null Hypothesis
Critics of ultrasound always cite its subjective nature as its great failing. Get over it. The subjective part is its strength and what differentiates this from an automated lab test. The user is an integral part of the system. Results will be best when the users are intelligent and well educated, motivated to improve, expert in equipment choice and operation, and adept at communicating and comforting patients.

A major dictum of a radiological education is that “normal” is the most difficult diagnosis. When have you looked enough and extracted enough information to know that nothing has been missed? We never know enough, yet we are impelled to act. This is the awesome responsibility of everyone who practices medicine. Operationally, normal is the null hypothesis that we are always trying to disprove. Classically, people are sent for imaging studies because of the likelihood of their having problems, so the null hypothesis is always before us. I suppose if you are dealing with an ambulatory population that is mostly normal to begin with and a noninvasive imaging technique, then, the medical null hypothesis might not be your prime directive. In situations where scanning and interpretation are severed operations, it is really had to cope with new and unfamiliar threats.