Hepatic surgery is based on anatomical principles. It is usually precise and can be nearly bloodless. Cure or palliation can be achieved for patients with unicentric primary or metastatic neoplasms, and even in some with multicentric malignancies.

The lungs are resectable, therefore, so is the liver. The liver and lungs are plethoric: afferent and efferent blood vessels intertwine. Each organ drains into a carina. In the lung the carina is at the bifurcation of the trachea. In the liver it is at the confluence of the bile ducts. During surgery certain conduits must be controlled in each of the organs. In the lungs these are the bronchii; in the liver, the bile ducts.

When haemostasis is perfect and cut bile ducts are closed or are drained into the small bowel, there is scarcely a limit to the extent of hepatic surgery, provided only that enough parenchyma is left to synthesize clotting factors, to detoxify, to conjugate metabolic products, and to perform the other functions of the liver.

While the lung, after resection, lies almost passive in the chest, cytologically speaking, the liver contains a population of long-lived normal cells that can be recruited to undergo mitosis when lost parenchyma must be replaced. Although the replacement of parenchymal mass is commonly termed regeneration, it is not. Regeneration means regrowth of an organ. This is seen in the ability of a newt to produce a new limb from the scar of an amputation stump, or in the annual sprouting of a stag's antlers. Hepatic regeneration is less complex, being characterized only by compensatory hypertrophy and hyperplasia of existing cells, which are mysteriously signalled to begin growth and are equally mysteriously stopped when the demand for new cells is met. Transforming growth factor-&bgr;&sub1;, a polypeptide, may be one mediator of these events. Hepatic growth factors also participate.

The anatomical importance of ‘regeneration’ is that new parenchyma grows only in the mass of existing parenchyma. After a right hepatectomy the patient's regenerative mass is centred on the remnant left liver, near the epigastrium; after a left hepatectomy the mass tends to be in the right hypochondrium.

The physiological importance of regeneration is the reserve that it supplies after a neoplasm is resected. As a neoplasm grows, it displaces normal liver tissue: it does not destroy it. In the unlikely event that parenchyma is damaged by the neoplasm, regeneration compensates. Enormous neoplasms can therefore be removed from the liver with little compromise of parenchymal mass. The same mass that was sustaining life before surgery is the parenchymal mass available for homeostasis afterwards.

The usual surgical mythology that Prometheus (‘forethought’) was chained to a rock for stealing fire from the gods is too simplistic. In actuality, Prometheus, son of the Titans Iapetus and Clymene, made man from clay, and stole fire from heaven (or from the forge of Hephaesteus) as revenge for Zeus's having deprived man of fire. Furthermore, when animal sacrifices were made, Prometheus tricked Zeus and other gods into choosing tough meat, while men got the better parts.

To punish Prometheus, Zeus had him chained to a rock in the Caucasus mountains, exposed to a vulture (or eagle), who feasted on Prometheus' liver by day, the liver regrowing nightly. (Perhaps this was the phase of his life in which Prometheus invented numbers.) Torture continued until Hercules freed Prometheus, or, in another version, until Prometheus revealed the secret of Thetis.

Prometheus knew that Thetis would have a son more powerful than his father, Zeus. Although Zeus loved Thetis, he knew also that he would be in danger from one of his sons, but not which one. Prometheus knew the secret, but he would not reveal it until he was freed, nor would he accept Pandora as a bride. Thetis was then married to Peleus, a mortal, so that her son would not be immortal. The son was Achilles.

The liver injured by trauma is less able to defend itself: the signal for the initiation of regeneration comes too late. Despite the belief that 90 per cent of the liver can be lost with impunity, the patient usually dies, from a combination of blood loss and coagulopathy. A patient who survives massive loss of parenchyma following resection of neoplasia is likely to experience chronic fatigue, since regeneration is finite.

The structural analogy between the lung and the liver proposed above is not just a pedogogic device, but practical knowledge: it ultimately determines both how to divide and resect the liver and how to identify functional units exactly.

A plane erected between the medial aspect of the gallbladder bed and the axis of the inferior vena cava splits the liver into two unequal parts.

Just as the right and left bronchi split at a carina to provide the architecture for moving air in and out of the right and left lungs, the liver may be divided bloodlessly, or nearly so, along the plane from the gallbladder bed to the vena cava, historically called Cantlie's line. Bile drains into the right and left hepatic ducts from the respective halves of the liver at a carina: namely, the confluence of the bile ducts. The right and left lungs are matched by the right and left liver. Just as a right or a left pneumonectomy is possible, so is a right or left hepatectomy. There are no true lobes in the liver. Thus, there can be no lobectomies, only hepatectomies and segmentectomies.

The liver is comprised of eight segments, each of which has both dedicated arteries and branches of the portal vein that nourish them and specific hepatic veins that drain them into the three major hepatic veins: right, middle, and left. (see Section 22.1 156) The nutrient sources of a segment are so invariable that during a hepatectomy a confident surgeon, in principle, does not need to ligate the bile duct, the portal vein branch, and the hepatic artery branch separately. Because the vessels are all within a single peritoneal investment (the Walaesian sheath), a single ligature around the contents of the sheath could be satisfactory. The small portion of liver lying on the intrahepatic portion of the inferior vena cava and draining direct to the vena cava is imprecisely, but probably irrevocably, called the caudate lobe (segment I). The imprecise term ‘trisegmentectomy’ recently introduced into surgical cant should be expunged.

With the exception of hepatic neoplasms such as carcinoid or glucagonoma that may secrete an unmistakable hormone, spot diagnoses of specific neoplasms are difficult; they were practically impossible before the advent of sensitive and specific means of imaging. Of all the methods of diagnostic imaging the most useful and economic is a CT scan enhanced by intravenous injection of a contrast medium.

The magnetic resonance image is an excellent adjunct, because the T&sub2;-weighted image uniquely displays haemangiomas and other tissues high in water content. MR is also the most sensitive way to detect metastatic disease in the liver.

Arteriography shows the arteries and delineates vascularized neoplasms. The experienced hepatic surgeon needs only to feel the arteries at the operating table to learn almost everything he needs to know about their distribution.

Exception to this argument is taken by some who support the need to identify and to treat in some special manner a replaced right hepatic artery or other arterial anomalies. These concerns usually are of little moment: a replaced right hepatic artery supplies the right liver and the gallbladder. Since these are highly likely to be the targets of surgery the replaced artery will have to be divided. Even if a replaced right hepatic artery were to be cut by accident, necrosis of the liver rarely develops, as long as portal vein dynamics are normal. The main advantages of arteriography are in disclosing vascular hepatic metastases, which are poorly detected by other means, and in illuminating the anatomy of neoplasms involving the inferior vena cava, the right renal vein, the extrahepatic portions of the hepatic veins, or the portal vein.

Intraoperative ultrasonography to identify hepatic metastases is rarely better than the experienced surgeon's hand. The principal exception concerns detection of hepatic cell carcinomas less than 4 cm in diameter, which are elusive without sonographic localization. Less importantly, sonography defines the route of intrahepatic veins and permits balloon occlusion and removal of a segment containing a small cancer in a cirrhotic liver.

Because hepatoadenomas are true neoplasms, they are supposed to lack Kupffer cells. In principle, they should not take up &sup9;&sup9;Tc&supm; sulphur-colloid. Focal nodular hyperplasia, not being a neoplasm, should be able to take up the colloid. Unfortunately, the specificity with which a differential diagnosis can be made is poor.

Knowledge of the histopathology of a neoplasm before surgery is only helpful when a neoplasm is suspected to be so disseminated that excisional surgery is useless, or when the liver is likely to be so diseased that the possibility of recovery is remote. Otherwise, focal hepatic neoplasms are routinely removed with a rim of normal liver tissue 1 or 2 cm thick, irrespective of the cell type.

Biopsy specimens are likely to be interpreted erroneously because of sampling error. If that were not the case, and if the possibility of encouraging peritoneal implantation of neoplastic cells or of precipitating haemorrhage from a haemangioma or from a vascular carcinoma could be dismissed, biopsy examination might theoretically be helpful in determining whether a tumour-like growth such as focal nodular hyperplasia or benign neoplasms such as hepatic cell adenoma could be spared. Imaging studies are generally safer and are specific enough for most purposes.

A right thoracoabdominal incision leaves nothing to chance, but causes unnecessary morbidity. Aside from the discomfort of a transected ensiform cartilage, closure of the diaphragm must be perfect to avoid bilothorax. A retrospective study of hepatic resections over a 27-year interval (1962–1988) indicated that a fall in the use of a thoracic incision from 57 per cent to 19 per cent was associated with a fall in the mortality rate from 19 per cent to 9.7 per cent; improved standards of care were obviously responsible for some of this decline. Resections of paracaval or of retrocaval neoplasms are practically the only indication for the thoracoabdominal approach. An abdominal approach is usually both adequate and feasible. Any requirement for more room, if the patient is short or fat, or both, or where the neoplasm is paracaval, is easily met by a modest median sternotomy. The patient's arm is suspended over his right chest wall to provide unlimited exposure. Nonetheless, an extensile abdominal exposure is required to take advantage of the elevated rib cage provided by a retractor fixed to the operating table. Much better mechanical advantage in elevating the thorax is provided by the bariatric surgery crossbar of a retractor such as the Omni-Tract&subR; than by that of simple, fixed retractors pulling in the plane of the abdominal wall. When a bilateral subcostal incision is used in conjunction with a ‘wishbone’ Omni-Tract retractor&subR;, exposure is not only less good, but closing the incision is more difficult because one of the elements to be sewn is cut muscle that is fixed to an unyielding rib cage. Moreover, a subcostal incision often heals with the contracting, caudad portion of the incision heaped above the level of the cephalad skin. At least four intercostal nerves on each side are also cut. The extensile exposure cuts only the intercostal nerves on one side, and the vertical line of the incision facilitates the use of both manual and mechanical retractors.

About half of a Finnish adult male population examined in forensic autopsies had benign liver neoplasms or tumour-like nodules, such as cholangioadenomas, focal nodular hyperplasia, and haemangiomas. The sizes of these neoplasms ranged from those that were only a few millimetres in diameter to those over 2 cm in diameter. There is no reason to suppose that other populations are different. Presumably, these tumours account for the unexpected lumps sometimes felt by meticulous surgeons who examine the liver routinely during a laparotomy for another reason. Such lumps should be ignored, unless an experienced surgeon regards one or more of them as being exophytic or umbilicated, and consequential.

Cholangioadenomas, also called bile-duct rests or hamartomas, are so common as to be ignored, even in postmortem examinations.

Although cavernous haemangiomas are frequent, they are rarely important. Those less than 8 cm in diameter seldom need treatment, except when an ill-advised biopsy provokes haemorrhage. Even then they generally bleed less than expected. Intervention is required only when the haemangioma (usually diagnosed best by MR imaging) causes symptoms of a mass lesion, displacing or irritating other organs, or if it ruptures. Leaks and ruptures are uncommon, however: fewer than 50 cases of ruptured haemangioma seem to have been recorded. All those that rupture are over 8 cm in diameter (about 270 cm³); a few smaller ones have leaked. Why one haemangioma 8 cm in diameter should leak, while another five times larger remains intact is a mystery. Perhaps the amount, the distribution, and the strength and compliance of the connective tissue are important considerations.

The common haemangioma is resectable along non-anatomical lines by any method favoured by the surgeon. Most, including giant haemangiomas, can and should be enucleated rather than resected, unless only a bridge of intact hepatic parenchyma has to be divided. There is no cause for concern if the resection margin of an enucleated or resected haemangioma contains evidence of haemangiomatous tissue, since they do not recur, except for rare instances in which concurrent oestrogen therapy potentiates growth.

These used to be regarded as a nosological curiosity. The initial paper showing that rupture of hepatic adenomas into the peritoneal cavity might be caused by the use of oral contraceptives was rejected, as being improbable.

Contemporary publicity has now alerted physicians and patients to the role of oral contraceptives in fostering the development of adenomas, although both pristine liver cell adenomas and the ruptured variety are infrequent. It is not yet determined whether hepatic adenomas progress to carcinomas. Although the risk is small, some adenomas, without doubt, undergo carcinomatous transformation.

Two-thirds of the liver parenchyma is right liver: most neoplasms, including adenomas (or colorectal metastases for that matter) are, therefore, located in the right liver. No one can predict which adenomas will rupture, nor why. When adenomas in the right liver burst, they are often deep-seated neoplasms, and an immediate hepatectomy, with its attendant risks, is sometimes required.

From a clinical viewpoint, any source of steroidal stimulation should be withdrawn from a hormone-treated man or woman thought to have a hepatic adenoma. If the neoplastic mass regresses by 50 per cent over the next 3 months, no treatment is likely to be needed. Adenomas that do not regress upon stopping hormonal therapy should be removed. Because of the possibility of carcinomatous transformation, albeit small, imaging studies at 3-month intervals for a total of 1 year should probably be pursued to assess the possibility of malignant growth.

This is a ‘condition’, not a disease or a neoplasm. It is probably a fibrogenic response to idiopathic ingrowth of blood vessels. If, because of its central scar and characteristic spherical or oblate shape the inferential diagnosis of a space-occupying mass is focal nodular hyperplasia, there is no indication for its removal, in the absence of symptoms. Concentration of &sup9;&sup9;Tc&supm;-labelled sulphur colloid in the tumour may help, or at least guide, diagnosis, but specificity is low: liver cell adenomas, formerly thought to be unable to concentrate the colloid, often do so.

Like focal nodular hyperplasia, cystadenoma is a tumour-like condition, not a neoplasm. Cystadenoma probably represents growth of a bud of bile-duct primordium. Fewer than 75 cystadenomas have been reported, the majority in women: they are four times more common in women than in men. Although hepatic cystadenomas produce all the discomfort and other effects of an intrahepatic mass, they do not cause jaundice unless they happen to compress a major bile duct. If symptoms persist, hepatic cystadenomas must be removed both for cure and to avoid confusion with the rare hepatic cystadenocarcinomas. Cystadenomas can usually be enucleated.

Hepatocellular carcinoma is by far the most common and important visceral neoplasm in the world, but it could be virtually eliminated by wide-scale inoculation against hepatitis B virus, provided that the level of protection afforded by the vaccine were high. The efficacy of the vaccine decreases in those over age 50 years, and people in third-world countries do not avail themselves of the vaccine, irrespective of their age. In transgenic mice injected with hepatitis B virus, intracellular progression of liver tissue from inflammation to carcinoma is unmistakable. Alcohol, aflatoxin, and male sex are other putative carcinogens or cocarcinogens.

Because multicentric hepatic cell carcinoma on a substrate of a liver ravaged by alcoholic cirrhosis is common in the Western world, resection is usually impractical, except for those who harbour encapsulated, unicentric cancers. Even though the best 5-year survival rate after resection is only about 35 per cent, aggressive surgical removal of a limited hepatic cell carcinoma is worthwhile, because there is otherwise no effective adjunctive therapy, although arteriographic chemotherapeutic embolization of the neoplasm may be beneficial.

The fibrolamellar variant hepatocellular carcinoma should also be treated surgically. This tumour is encapsulated and occurs mostly in women (but sometimes also in steroid-treated men); it can usually be resected, with a survival rate as high as 40 per cent after 4 years.

Minimal hepatocellular carcinomas (those <4 cm in maximal diameter) are rare in the United States, but common in the ‘hepatic cancer belt’ of Africa and Asia and in France also, as a result of historical referral patterns. In the ‘cancer belt’, screening programmes identify these neoplasms as a matter of routine: first, by assaying for abnormally high levels of &agr;-fetoprotein in the blood and, second, if those are high, by ultrasonographic scans of the liver.

Even in patients with considerable cirrhosis, hepatitis, and hepatic dysfunction, minimal cancers can be removed by a bloodless resection of the neoplasm with a few millimetres of apparently normal parenchyma as described above. Asanguineous conditions are met either by occlusion of portal blood flow from the hepatic segment harbouring the neoplasm or by total occlusion of blood flow to the liver.

Because hepatocellular carcinomas invade the bile ducts and the hepatic and portal veins, deposits of cancer in the bile ducts may produce obstructive jaundice. Neoplasm in the hepatic and portal veins may cause portal hypertension and Budd-Chiari syndrome, manifested by bleeding oesophageal varices.

This is a rare, idiopathic, and limited cancer which is sometimes curable by resection. By analogy with hepatic cystadenomas, cystadenocarcinomas are thought to originate from bile duct primordia.

Cholangiocarcinomas probably originate from cells in the lining of bile ducts; the entity of biliary papillomatosis may represent one of its stages. Cholangiocarcinomas mimic some of the behaviour of hepatocellular carcinomas in that they may coexist in a parenchymal (or ‘peripheral’) phase and a ‘central’ phase, invading bile ducts widely and producing symptoms of bile duct obstruction as the chief manifestation of disease.

Bile ducts can sometimes be freed of cancer, and regions of the biliary system not amenable to ‘thrombectomy’ of the neoplasm can be resected, yielding palliation of symptoms for several years. Parenchymal multicentricity is normally inexorable; however, one of our patients is alive 4 years after resection of all the hepatic segments except segments I, II, and part of segment V, combined with thrombectomy of the ducts from segments V and VI, which were drained into a Roux-en-Y loop.

Angiosarcomas are interesting because they are rare and incurable, but mainly because their aetiology is usually specific and obvious. They usually arise following injection of Thorotrast decades ago (a radioactive imaging substance that emits thorium daughters), following arsenic exposure of agricultural workers, especially in the vineyards of the Moselle Valley, following testosterone in boys with Fanconi's syndrome, and after exposure to vinyl chloride (the monomer of many plastics). Sarcomas otherwise are rare.

Although angiosarcoma is incurable, leimyosarcomas and embryonal rhabdomyocarcomas may be curable by excision and adjuvant chemotherapy. Because some fibrosarcomas and leiomyosarcomas secrete an insulin-like growth factor, they initially manifest themselves by causing hypoglycaemia.

Countless metastases lodge in the liver: few grow, and only the minority are worth removing. Nevertheless, huge, indolent metastases from exotic sources, such as mesonephric duct carcinoma, are sometimes resectable with the expectation of cure or of long-term palliation.

Colorectal primary cancers contribute the greatest number of resectable metastases to the liver; these are often indolent. Metastases originating in the right colon tend to be myriad and biologically aggressive; those from the descending colon and rectum are more likely to be solitary, or at least enumerable. An estimate has been made that only seven of 1750 patients with untreated hepatic metastases from the colon and rectum have ever survived 5 years.

By removing metastases the hypothesis is that the patient will be shifted from membership of a group about to die into a group that has a 25 to 35 per cent 5-year survival rate. However, no one is certain of the natural history of untreated colorectal metastases. Some remain static for years, having a growth period of 2 to 3 years before they become obvious.

Resection of four or fewer metastases is appropriate, including any satellites clustered around the central neoplasm. Resection of more than four metastases is generally futile, because with that degree of metastatic spread, both deposits in the liver and occult lymphatic and vascular metastases are often widespread.

The hope that removal of an enormous mass of metastatic colorectal cancer by a major hepatectomy will relieve pain and discomfort is often unrealized.

Chisari FV, et al. Molecular pathogenesis of hepatocellular carcinoma in hepatitis B virus transgenic mice. Cell, 1989; 59: 1145–56.

Franco D. Malignancy of the liver. Curr Opin Gastroenterol, 1990; 6: 447–53.

Franco D, et al. Resection of hepatocellular carcinomas: results in 72 European patients with cirrhosis. Gastroenterology, 1990; 98: 733–8.

Ishak KG, Malt RA. Sarcomas of the liver and spleen. In: Raaf JH, ed. Management of Soft Tissue Sarcomas. Chicago: Year Book Medical Publishers 1992: 165–80.

Karhunen PJ. Benign hepatic tumours and tumour-like conditions in men. J Clin Pathol, 1986; 39: 183–8.

Savage AP, Malt RA. Elective and emergency hepatic resection: determinants of operative mortality and morbidity. Ann Surg, 1991; 214: 689–95.

Stocker JT, Ishak KG. Focal nodular hyperplasia of the liver: a study of 21 pediatric cases. Cancer, 1981; 48: 336–45.

Takayama T, et al. Malignant transformation of adenomatous hyperplasia to hepatocellular carcinoma. Lancet, 1990; ii: 1150–3.

Wanless IR, Mawdsley C, Adams R. On the pathogenesis of focal nodular hyperplasia of the liver. Hepatology, 1985; 5: 1194–200.