EVIDENCE-BASED CLINICAL EXAMINATION  —  CHAPTER 3

The Abdominal Examination

From Practice to Evidence · Which of its signs survive a second examiner?

Castell’s signMurphy’s signPocket ultrasound

The chest examination, in the previous chapter, turned out to have one sign that two examiners almost always agree on, asymmetric expansion, and a long tail of signs they do not. The abdomen is a harder case. Its organs are deeper, its wall is thicker and more variable, and its signs were named for the surgeons and physicians who described them in single patients a century ago. Murphy described his sign in 1903, McBurney his point in 1889, Castell his percussion sign in 1967, and each name has been taught since as if the naming settled the question. The rational examination asks the same two things of each: does finding it change the probability of disease, and would a second examiner have found it?

What follows is the examination as it should be performed, and then the answers, organ by organ.

The examination as it should be performed

Setting up. The patient lies flat with one pillow, arms at the sides, the abdomen exposed from the xiphisternum to the symphysis pubis, in good light and a warm room. Examine from the right. Ask where it hurts before touching, and examine that quadrant last. Before the hands go on, spend thirty seconds watching: whether the patient lies still, as in peritonitis, or moves restlessly, as in colic; the movement of the abdominal wall with breathing; and what is at the bedside.

The peripheral survey.

Inspection of the abdomen. Contour, from the side as well as from above: distension, and whether it is generalised or in the flanks. Scars, named and dated. Distended veins and the direction of their flow. Striae, bruising in the flanks or around the umbilicus, visible peristalsis, visible pulsation, hernial orifices with a cough, and the umbilicus itself.

Palpation.

Percussion. The liver span in the right mid-clavicular line, from resonance above to dullness and from tympany below; the spleen by Castell’s sign, percussion in the lowest intercostal space in the left anterior axillary line in expiration and full inspiration, and by Traube’s space; the bladder; and, for fluid, flank dullness, shifting dullness with the patient rolled and the note re-taken after thirty seconds, and the fluid wave with an assistant’s hand on the midline.

Auscultation. Bowel sounds for at least a minute before calling them absent; bruits over the aorta, the renal arteries and the liver; a venous hum; a friction rub.

Completing the examination. The hernial orifices standing, the external genitalia, a digital rectal examination, the legs for oedema, the urine, and a temperature chart. Then, as with the chest, the findings are put into a pattern, an acute abdomen, portal hypertension, an obstructed bowel, a mass with its likely organ, rather than left as a list of signs.

Fluid in the abdomen

Ascites is the best-studied problem, and its story is the story of the whole examination in small. In 1982 Cattau and colleagues, Donald Castell of the percussion sign among them, examined twenty-one patients referred with questionable ascites and checked the signs against ultrasound. Six of the twenty-one had fluid. The sensitivity of the individual manoeuvres ranged from 50 to 94 per cent and the specificity from 29 to 82, and the overall accuracy was 58 per cent. Only one statement could be made with better than 90 per cent accuracy: that if there is no flank dullness there is no ascites. The authors recommended ultrasound for any questionable case.

Simel and colleagues in 1988 put numbers on the whole bedside assessment in a broad range of inpatients. When the examiner judged ascites likely, the likelihood ratio was very high, between 38 and 83; when the examiner judged it unlikely, the ratio was only 0.8 to 0.9, that is, an unimpressed examiner had barely lowered the probability at all. Williams and Simel drew the lesson in the JAMA series in 1992: the fluid wave is the most specific sign and the least sensitive, while bulging flanks, flank dullness and shifting dullness are sensitive and not specific. Shifting dullness, at about 83 per cent sensitivity and 56 per cent specificity for volumes above a litre and a half, is the sign to rely on for exclusion; the fluid wave, and peripheral oedema, are the findings that raise the probability most; and the most useful things for ruling ascites out are two questions, whether the ankles have swollen and whether the belt has been let out.

The liver and the spleen

Naylor’s review of the liver examination for the same series in 1994 gave the palpable liver edge a positive likelihood ratio of about 2.5 for hepatomegaly and a negative ratio of about 0.45, which is to say that a liver edge below the costal margin raises the probability of an enlarged liver a little and its absence lowers it a little, and that neither settles anything. The liver span by percussion fared worse: across the studies, the proportion of spans within two centimetres of the imaging measurement ranged from 38 to 78 per cent, and Sullivan, Krasner and Williams had shown in 1976 how much of that error comes from the examiner rather than the liver. Joshi and colleagues repeated the exercise in 2004 in a rural Indian hospital, three physicians examining 180 patients blind to each other and to the ultrasound, and found palpation and percussion each of modest accuracy and modest agreement.

The spleen is the organ for which the examination has been studied most carefully, and the results are the most instructive. Grover, Barkun and Sackett reviewed the evidence in 1993. Castell’s percussion sign was the most sensitive manoeuvre, at about 79 per cent, but with a specificity of 46 per cent; Traube’s space gave 62 and 72; and palpation was specific and insensitive. The combination did what neither did alone: when percussion and palpation were both positive, the sensitivity fell to 46 per cent but the specificity rose to 97, and palpation turned out to discriminate only in patients whose Traube’s space was already dull. From this came the sequence still taught: percuss first, and palpate only if the percussion is dull.

But the same year Tamayo and colleagues asked the question the review could not: does it depend on who is examining? It did. Across their examiners the sensitivity of palpation ranged from 0 to 64 per cent, and of percussion from 8 to 75. A sign whose sensitivity depends on the examiner more than on the spleen is not yet a test, however good its pooled numbers look.

The acute abdomen

For the painful abdomen the question is not size but decision, and the eponyms were tested as decisions. Trowbridge, Rutkowski and Shojania reviewed acute cholecystitis in 2003 and found that no single item of history, examination or basic laboratory testing could confirm or exclude it; the nearest was Murphy’s sign, with a positive likelihood ratio of 2.8 and a confidence interval that reached from 0.8 to 8.6, that is, one that included the possibility of no value at all. Wagner, McKinney and Carpenter had reviewed appendicitis in 1996: right lower quadrant pain carried a positive likelihood ratio of about 7 to 8, rigidity about 3.8, migration of the pain about 3.2, and the psoas sign about 2, with fever and vomiting before the pain adding their smaller shifts. None is decisive; together, in a score such as Alvarado’s, they set the pre-test probability that decides whether the patient goes to the scanner.

Lederle and Simel, in 1999, asked whether palpation finds an abdominal aortic aneurysm. Pooling fifteen screening studies in which patients had both palpation and ultrasound, they found that palpation was the only manoeuvre of any value, and that its sensitivity depended on the size of the aneurysm: 29 per cent for aneurysms of three to four centimetres, 50 per cent for four to five, and 76 per cent for five and above. The hand finds the aneurysm that most needs finding about three times in four, and misses the small one that could have been followed.

And here reproducibility was measured directly, in the setting where it matters most. Pines and colleagues in 2005 compared attending and resident physicians examining the same emergency patients with abdominal pain, 122 pairs in all. Agreement on the presence of a mass was almost perfect; on guarding, distension and tenderness it was only moderate. Yen and colleagues, the same year, compared residents, attendings and surgeons examining children with abdominal pain, and found chance-adjusted agreement below moderate on almost every component, with kappa values from −0.04 to 0.38 between residents and attendings; the one finding on which attendings and surgeons agreed moderately was rebound tenderness, at 0.54. The sign that decides whether a child goes to theatre is one that two experienced examiners agree on little better than half the time beyond chance.

The sounds of the bowel

Bowel sounds are the abdominal equivalent of the whispered voice: taught to everyone, recorded in every note, and, when tested, nearly without value. Felder and colleagues in 2014 played recordings from normal patients, from patients with mechanical small-bowel obstruction and from patients with post-operative ileus to clinicians, and found positive predictive values of 23, 28 and 44 per cent, and that the same clinician, hearing the same recording twice, gave the same answer only 52 to 59 per cent of the time. Breum and colleagues in 2015 had fifty-three doctors listen to recordings from ninety-eight patients with suspected obstruction, thirty-five of whom proved to have it: the median sensitivity was 42 per cent, the median specificity 78, and the median kappa across 1,378 pairs of doctors was 0.29. The tinkling bowel sound of obstruction, in other words, is heard by some doctors and not others, and by the same doctor on some days and not others.

What has changed since 2020

Ultrasound has arrived at the bedside, and for the abdomen it has largely won. The evidence is older here than for the chest, and stronger. Rubano’s systematic review of emergency-department ultrasound for aortic aneurysm found a pooled sensitivity of 99 per cent and specificity of 98, against the 29 to 76 per cent of the hand. Ross’s review of emergency-physician ultrasound for gallstones, eight studies and 710 patients, found sensitivity of about 90 per cent and specificity of 88. Gottlieb’s meta-analysis of eleven studies found ultrasound for small-bowel obstruction 92 per cent sensitive and 97 per cent specific, against a stethoscope that finds it 42 times in a hundred. Keil-Ríos and colleagues in Mexico City found a pocket device 96 per cent sensitive for ascites. Appendicitis is the exception that proves the pattern: Fields’ meta-analysis of 2017 gave point-of-care ultrasound 91 per cent sensitivity and 97 per cent specificity overall, but only 80 and 92 when emergency physicians rather than radiologists held the probe, and Becker’s multicentre study of 2022 concluded that in an undifferentiated emergency population it is moderately accurate and not a definitive test. Ultrasound is also operator-dependent; it is simply less so than the hand.

The recorded bowel sound is the second front, and it is younger. Once the stethoscope became a microphone, the sound could be stored, timed and classified. Nowak and colleagues reviewed automated bowel-sound analysis in 2021, and a benchmark of 2025 compared feature-based and pretrained models on recorded bowel-sound patterns and found the pretrained models better, particularly on the rarer patterns. The aim so far is the long-term, non-invasive monitoring of motility, in ileus, in inflammatory bowel disease, after surgery; nothing yet replaces the examiner at the bedside, and the datasets are small.

The examiner is still the variable. Arora and colleagues in 2017 randomised first-year students to a brief teaching in either ultrasound or physical examination for hepatomegaly and ascites and found no difference in their reliability or accuracy, though the students preferred the probe. In inexperienced hands the two tools were equally poor, which is a point about training rather than about tools.

What the numbers should change in teaching

Castell was a co-author of the study that showed his own region of the examination to be 58 per cent accurate, and recommended ultrasound. That is the right relation between a physician and an eponym: the name marks where the hand once had to go alone, not where it must stay.

✦ ✦ ✦

Sources

© 2026 Husain Alkhaldy — Evidence-Based Clinical Examination.