1. Background

The purpose of this paper is to help clinicians and patients make decisions about the management of gallstone disease that is either asymptomatic or has caused only biliary pain. Management of the complications of gallstone disease, such as acute cholecystitis and cancer of the gallbladder, is not included. We describe the clinical features that help to differentiate symptomatic disease from asymptomatic (silent) gallstones and then compare their natural histories. We discuss the efficacy, safety, and benefits of surgical and nonsurgical methods for treating gallstones and compare the mortality risks of immediate surgery with the risks of expectant management.

2. Gallstone Disease

2.1. Background

Gallstones are very common. By age 75, about 35% of women and 20% of men have developed gallstones [1-3]. Most of these persons are asymptomatic and unaware they have gallstones. If a biliary problem does occur, it is usually an episode of biliary pain. Less common problems but much more serious are the acute and chronic complications of gallstone disease (Table 1). Death caused by gallstones is uncommon [4, 5] and accounts for roughly 5000 of the 2.2 million deaths per year in the United States. Most gallstone-related deaths occur in the elderly and are caused by biliary complications, including the surgery done to treat complications.

2.2. Symptoms

Symptoms are almost always episodic and infrequent [6, 7]. An attack of biliary pain typically develops rapidly and is severe, steady, and unrelieved by usual household remedies, position change, and gas passage. If pain episodes do recur after the initial presentation, the frequency may vary from weeks to years. A first episode of biliary pain may be frightening until it is evident that neither a heart attack nor an abdominal catastrophe has occurred. The clinician should not attribute to gallstones pain that is present much of the time, that frequently comes and goes, that lasts less than 15 minutes, or that is not severe enough to interfere with activity. Nor should the clinician attribute to gallstones those symptoms of dyspepsia Table 1 that occur as often in persons without stones as in those with stones [8-10]. The appropriate attribution of symptoms is important when making management decisions because the expected prognosis of symptomatic gallstones is much different than that of asymptomatic gallstones.

Persons with gallstones do not have a "chronic disease" in the sense that they do not have the long-term disability that accompanies such diseases as congestive heart failure, stroke, and cancer. Thus, the decision for the physician and patient is not whether to alleviate a chronic or debilitating condition but rather is whether to prevent (for example, by cholecystectomy) a potential future problem, either biliary pain or a biliary complication. To make this decision requires information about prognosis for these outcomes and about the efficacy, safety, and effort of possible therapies.

2.3. Diagnosis of Gallstones

Diagnosis of the presence of gallstones is relatively simple and can be assessed by ultrasound or, less commonly now, by oral cholecystography or plain roentgenography [11]. The determination of whether a patient's symptoms are attributable to gallstones represents an important diagnostic challenge. Laboratory abnormalities, such as those caused by biliary complications, are uncommon in persons with simple biliary pain. Thus the diagnosis of whether gallstones are symptomatic is based primarily on history. Asymptomatic gallstones are usually discovered incidentally during ultrasonographic examination or laparotomy for nonbiliary abdominal problems.

2.4. Therapeutic Approaches to Gallstones

There are three basic therapeutic approaches: surgical, nonsurgical, and expectant management. The surgical approach includes open cholecystectomy and laparoscopic cholecystectomy. The nonsurgical approach involves removal of gallstones but not the gallbladder. In a "wait-and-see" approach, intervention is delayed until a problem later develops. This approach, which we call "expectant management," involves the trade-off between the benefit of possibly avoiding an intervention that may never be needed versus the risk that, if an intervention is required at an older age, the associated mortality and morbidity is usually increased. Thus, if a serious complication like acute cholecystitis or pancreatitis requires urgent surgery at an older age, the mortality rate will be substantially higher than if elective surgery had been done at an earlier age.

Surgical removal of gallstones and the gallbladder prevents future episodes of biliary pain or complications. Removal of the gallbladder is generally not associated with adverse effects, although some cases of postcholecystectomy diarrhea have been reported [12]. The "post-cholecystectomy syndrome" may refer to the "recurrence" of symptoms that preceded the cholecystectomy and were not caused by gallstones in the first place [13]; the term has also been used to refer to a pain similar to biliary colic and that may be due to "biliary dyskinesia" [14]. The clinical challenge in decision making is to balance the projected prognosis of expectant management versus the risk, effort, inconvenience, and cost of an immediate intervention.

3. Methods

We did literature searches using bibliographies from reviews and recent articles to acquire reports about the safety and efficacy of different treatment methods and about the natural history of persons with gallstones. We used a MEDLINE search to find English-language articles about laparoscopic cholecystectomy. We then used simulation modeling to project expected outcomes for expectant and operative treatment strategies. We consider nonsurgical treatments in a qualitative manner. The model is intended to apply to persons of various ages with usual co-existing diseases for that age; the comorbid conditions commonly associated with increasing age are accounted for in the mortality rates used for operation and for death from other causes.

4. Results

4.1. Treatment Methods

Both surgical and nonsurgical treatment methods are available. Compared with nonsurgical approaches, surgical methods have several advantages: 1) wide applicability to gallstones of any number, size, and composition; 2) prompt removal of stones, in contrast to the sometimes lengthy treatment required to dissolve gallstones using nonsurgical methods; 3) prevention of gallstone recurrence; and 4) prevention of gallbladder cancer because the gallbladder as well as gallstones are removed. The major disadvantages of surgery include: 1) the risk for death which, although not great in an absolute sense, is higher than that for nonsurgical therapy; and 2) the inconvenience of surgery which entails time lost from work and other activities. One advantage of laparoscopic cholecystectomy is the reduction of that inconvenience.

4.1.1. Surgical Methods

Open cholecystectomy has for a century been the standard method to treat gallstones and is one of the safest of surgical procedures. Both open cholecystectomy and laparoscopic cholecystectomy require general anesthesia and incur its associated risks. A large number of studies have reported operative mortality rates for open cholecystectomy from both community and university hospital settings. The mortality rates shown in Table 2 are based on 12 reports concerning over 725 000 persons from 1950 to 1978. Most of the persons were in the two studies of the Commission on Professional and Hospital Activities and in the National Halothane Study, which involved over 1500 hospitals [15-17]. The method for compiling the data has been previously described [18]. Subsequently, one large study has reported generally similar operative mortality rates [19]. Two recent but smaller studies suggest that the mortality rate for open cholecystectomy may have declined over the last decade [20, 21].

The assessment of operative morbidity is problematic because of the wide variability in detail in reporting frequency, type, and severity of morbid events. We have not tried to summarize morbidity rates here. The most serious complication is common bile duct injury, which may require difficult reparative surgery and repeated hospitalizations and may ultimately be fatal; bile duct injury is uncommon and occurs in approximately 0.1% to 0.2% of cases [22, 23].

The data in Table 2 show that open cholecystectomy has, in general, low mortality rates. The mortality rates for women are approximately half those for men, and the mortality rates for nonelective or urgent cholecystectomy are approximately four times higher than for elective cholecystectomy.

One important practical feature of open cholecystectomy is that, because it requires a more extensive incision including the rectus abdominus muscle, a patient's recovery time to full activity may take months. The shorter recovery time for laparoscopic cholecystectomy is an important advantage.

Laparoscopic cholecystectomy [24-32] has been available since 1989 and has already gained wide popularity among both surgeons and patients. The gallbladder is approached usually through four small incisions or ports through the abdominal wall so that the rectus muscle is not incised. Conversion to an open cholecystectomy is needed in approximately 5% of cases. The safety of laparoscopic cholecystectomy has not yet been reported from very large studies. The results of currently available studies are listed in Table 3.

The data in Table 3 regarding the safety of laparoscopic cholecystectomy must be interpreted with caution. Because relatively few procedures have yet been done in contrast to open cholecystectomy, precise estimates of mortality cannot be made either in aggregate or stratified by a patient's age. Second, the rates have not yet "stabilized" because not all surgeons have fully trained in the procedure and learned how to prevent important complications such as bile duct injury or bowel perforation. Thus current mortality rates may be overestimates which reflect a learning curve. On the other hand, because only small and selected experiences have thus far been reported, in contrast to the large and comprehensive assessments of open cholecystectomy, the currently reported rates may be underestimated. There is some evidence that, at least currently, the surgical complication rates for bile duct injury may be substantially higher for laparoscopic cholecystectomy than for open cholecystectomy [33, 34]. Some experienced surgeons believe that the laparoscopic approach will eventually prove to be as safe or even safer than open cholecystectomy because pulmonary and cardiac complications may be fewer and because the current causes of common duct injury should eventually become understood and minimized. Only time will tell how safe the new procedure can be.

4.1.2. Nonsurgical Methods

Although nonsurgical methods have not become popular, we describe them here. We will not consider their use quantitatively in simulation modeling, but we will discuss their use qualitatively in later sections.

Oral bile acids act to dissolve cholesterol stones by increasing the solubility of cholesterol in bile. This approach is generally applicable only for gallstones made up primarily of cholesterol and of a diameter of less than 1.5 cm. Approximately 20% of persons undergoing cholecystectomy meet these requirements and would be eligible for treatment with oral bile acids [35]. The best candidate would have stones that are under 0.5 cm in diameter and float during oral cholecystography [36]. The efficacy of therapy depends on the size and cholesterol content of stones and on patient compliance; the overall dissolution rate in a 2-year period may range from about 25% [37] to 50% [38] and is higher for "floating" stones, about 67%, compared with a rate of 32% for "nonfloating stones" [36]. The efficacy of bile acid therapy may be enhanced by appropriate selection of patients [39]. Also, gallstones that are newly formed, for example, after weight reduction, pregnancy, or therapy with some lipid-lowering drugs, may be particularly amenable to dissolution with oral bile acids. Bile acids must be taken daily for up to 2 years. Bile acid therapy has not become popular in the United States, presumably because of limited efficacy and inconvenience, and because it is being superseded by other methods.

Methyl-tert-butyl-ether, a powerful contact solvent, acts by dissolving cholesterol during repeated instillations into the gallbladder. An interventional radiologist passes a catheter percutaneously through the liver into the gallbladder. Frequent instillations and removal of solvent take 12 hours or more. Stones of any size or number may be successfully treated with methyl-tert-butyl-ether. The degree of dissolution was reported to be 95% to 100% complete in 72 of 75 patients at the Mayo Clinic [40]. Complications may occur if the ether escapes into the peritoneal cavity or via the cystic duct into the small intestine. This procedure is investigational and is currently being done in only a few centers.

Extracorporeal shock wave lithotripsy uses acoustic shock waves to break stones into small pieces, which, after additional bile acid dissolution therapy, may dissolve or pass into the small intestine. Shock waves are generated by spark-gap, electromagnetic, or piezoelectric technology, and the procedure takes roughly 1 hour. Stones should be radiolucent and have a combined (or additive) diameter no greater than 3 cm. Approximately 12% of patients having surgery for gallstones would have a solitary radiolucent stone under 2 cm in diameter and be the best candidates for lithotripsy [35].

After lithotripsy and oral bile acid therapy, approximately 60% of persons with a solitary 2-cm stone may have complete clearing within 6 months and 90% within 1 year [41, 42], although success rates may be lower [43]. After lithotripsy, an episode of pain occurs in about one third of patients [44], but biliary complications are rare. Although this procedure received great interest in the late 1980s based on studies from Europe, it has not been approved by the Food and Drug Administration for general use in the United States and is now being superseded by laparoscopic cholecystectomy. Internationally, however, an increase in delivery of shock-wave energy and recognition of the importance of fragment size have improved efficacy and suggest that there may be a future role for lithotripsy in selected patients, at least in tertiary medical centers [43].

Among nonsurgical approaches, oral bile acids have limited applicability (that is, primarily to small radiolucent stones), and treatment generally must be continued for many months. Methyl-tert-butyl-ether can be applied to a wider spectrum of stones and provides rapid dissolution. Persons treated with lithotripsy may need months of oral bile acid therapy to achieve clearance. One potential disadvantage of all three nonsurgical methods is the effort of subsequent follow-up to detect stone recurrence. Recurrence is expected in about 10% of persons each year for about 5 years after initial dissolution [45]. Some evidence suggests that the recurrence rate reaches a plateau at that point [46, 47],_but other evidence suggests that recurrence may happen even beyond 5 years [48]. Recurrence is more common in persons with multiple stones [48]. It is still unclear whether recurrent stones need additional treatment and, therefore, whether surveillance is needed to detect and treat them. That decision would be based on whether the natural history of recurrent stones is similar to that of the originally treated stones [49]. Another potentially important consideration for nonsurgical therapy is that it may not prevent gallbladder cancer if the occurrence of cancer is related not to gallstones but to properties of bile or of the gallbladder.

4.2. Natural History

4.2.1. Asymptomatic Gallstones

Once they have formed, gallstones rarely disappear spontaneously. An exception is that stones formed during the metabolic disturbances that accompany weight reduction [50] or pregnancy may spontaneously disappear [51].

Once present, gallstones may lead to biliary pain or complications. The natural history of asymptomatic gallstones (that is, the rates to develop subsequent biliary pain or a biliary complication) is described in articles listed in Table 4 [52–63]. In these studies, gallstones were discovered for some reason other than typical biliary-tract pain. Only two studies [56, 57] were both of long duration and presented data using actuarial analysis.

We have previously suggested that the rate to develop biliary pain is about 2% per year for 5 years and then may decrease over time [57]. In that study a biliary complication occurred in 3 of 123 asymptomatic persons, and in all 3 an episode of pain preceded the complication and thus provided a "warning" symptom that might have allowed an intervention before the complication occurred. One limitation of this study is that most of the participants were white men, although a study involving women reports similar results [62]. Several studies report that the rate of subsequent symptoms appears to decrease over time [52, 56, 58, 62]. One possible reason for this apparent decrease may be that "younger" stones tend to cause more symptoms; alternatively, perhaps patients report symptoms more readily soon after learning they have gallstones.

Friedman and colleagues [56] report a rate of 3% to 4% per year for "any (biliary) event" during the first 10 years of follow-up. These events consist of biliary complications, biliary pain, and "cholecystectomy for chronic or milder symptoms." The rate decreased to 1% to 2% per year during the next 10 years. Data from Newman and colleagues, although not presented in life-table form, can be interpreted to show a 19% cumulative probability to develop pain over 10 years or approximately 2.2% per year [64]. A more detailed discussion of natural history is in Appendix 1.

One important feature in assessing biliary pain as an outcome is that episodes of pain, when they do occur, are usually of short duration and are isolated or infrequent and thus do not represent a drastic clinical outcome such as death or major disability. Because biliary complications are such an important outcome and can be fatal, the mode of presentation of biliary problems needs assessment. If the presentation is typical biliary tract pain, then elective treatment such as cholecystectomy may be done with a lower mortality rate than if the presenting problem is a biliary complication. Several studies suggest that when persons with asymptomatic gallstones develop a biliary problem, pain, not a complication, is the presenting problem in about 90% of cases [52, 57, 62-64].

About one third to one half of all the gallstone-related deaths in the United States are caused by gallbladder cancer. The median age of persons with gallbladder cancer is 73 years [65]. Although gallstones are present in about 80% of persons with gallbladder cancer, it is uncertain whether gallstones play a causal role; an alternate explanation for the association is that a common underlying feature such as property of the bile, is responsible for both cancer and gallstones. If cancer is related to bile and not to gallstones, then nonsurgical therapies might have no influence on the development of gallbladder cancer because the gallbladder is left intact.

In certain subgroups of persons (for example in Pima Indians), the lifetime risk for gallbladder cancer is high [66]. Pima Indians also develop gallstones at a young age [67]. It is unclear, however, whether the high incidence of cancer is related primarily to the age of the gallstone or to the age of the gallbladder (that is, the patient's age) and presumably to properties of bile. The answer to this question may have important implications for the expected lifetime risk of cancer, particularly for young persons with gallstones. The derivation of rates of gallbladder cancer can be based on either age of the patient or on known duration of gallstone disease. The different approaches yield different results and may have important implications for natural history and therefore for treatment. These issues are discussed in more detail in Appendix 1 and in the results section.

4.2.2. Symptomatic Gallstone Disease

The natural history of symptomatic gallstone disease, that is, of persons who have had an uncomplicated episode of biliary pain, is described in the articles listed in Table 5 [7, 56, 58, 61, 68, 69]. Only two studies present data in life-table analysis form [7, 56].

The recurrence of biliary pain in the National Cooperative Gallstone Study was about 38% per year over 2 years in persons who had had pain within 1 year before entry into this study and about 17% per year over 2 years in persons who did not have pain within 1 year before entry. (The presence or absence of symptoms before the year before entry is not described.) Most of the patients with recurrent pain had only a single episode during the 2 years of follow-up. Newman and colleagues similarly reported that about 50% of persons had recurrent pain within 1 year [64]. They also observed that about 30% of persons had no recurrence of pain during a 10-year follow-up. This study, along with others [52, 55, 58, 62, 68], suggests that a subgroup of persons may have one or two episodes of biliary pain and then remain free of pain for many years. Newman and colleagues [64] reported a "distinct impression that the initial pattern (of biliary pain) is preserved" so that patients seemed to have either relatively frequent episodes or widely spaced episodes.

The ascertainment of a biliary complication is likely to be less problematic than ascertainment of pain because complications are diagnosed objectively and generally lead to hospitalization. Several studies [7, 56, 59, 64] indicate that the rate to develop a biliary complication is 1% to 2% per year and that the rate remains steady for many years [56, 64].

4.2.3. Natural History after a Biliary Complication

After an episode of acute cholecystitis or gallstone pancreatitis, a person not treated by cholecystectomy has roughly a 30% chance over 3 months of having a recurrent episode [70, 71]. For this reason, persons who have had a complication of gallstone disease are considered to need early treatment and are not discussed further here.

4.3. Projected Outcomes of Different Treatment Strategies

By using simple mathematical modeling techniques described previously ([72] as modified from [18]), the long-term outcomes of persons with asymptomatic and symptomatic gallstones were projected for a strategy of expectant management and were contrasted with results for other treatment strategies.

In structure, two basic models are used. Figure 1 shows a person's possible "transitions" from having no gallstones to having gallstones and their sequelae. For asymptomatic gallstones, possible transitions include those marked with an "A" to continue to have asymptomatic gallstones; to develop biliary colic and have an elective operation (and possibly to die from the operation); to develop an acute biliary complication and have an urgent operation (and possibly to die from the operation); to develop gallbladder cancer and to die; and to die from other causes. For symptomatic gallstones, possible transitions include those marked with an "S." The assumptions and transition probabilities used are described in Appendix 1.

View larger version (26K): [in this window] [in a new window]   Figure 1. The model can be represented as a Markov state problem with several states. For calculations in the simulation model, persons with asymptomatic gallstones could have transitions marked "A"; persons with symptomatic gallstones could have transitions marked "S." For details, see text.

4.3.1. Asymptomatic Gallstones

4.3.1.1. Open Cholecystectomy. The results of expectant management were projected for hypothetic groups of 30-year-old, 50-year-old, and 70-year-old men and women and were compared with prophylactic cholecystectomy using mortality rates for open cholecystectomy. Table 6 indicates the cumulative risk of death from gallstones if asymptomatic gallstones were managed expectantly with no therapy until biliary pain or a biliary complication occurs. In expectant management, an elective cholecystectomy is done after an episode of biliary pain, and an urgent cholecystectomy is done if the presenting problem is a complication like acute cholecystitis. For a complication, the operative mortality rate is higher than for elective cholecystectomy.

Table 7 shows the average life expectancy that would be lost by persons using expectant management compared with persons who have a prophylactic cholecystectomy when gallstones are discovered. Discounting is shown to indicate that a day of life expectancy in the distant future has less value than the same day lost in the near future [73].

Base Case Analysis. For persons with asymptomatic gallstones managed by expectant management, the probability of ever needing a biliary operation is roughly 30% for 30 year olds, 20% for 50 year olds, and 15% for 70 year olds. The reason that women have a slightly higher lifetime probability of needing an operation compared with men is that women generally live longer and have a longer time to develop a biliary problem. The cumulative risk for gallstone-related death for women is lower than for men because operative mortality rates are lower for women than for men. As shown in Table 6, column labeled [1], for younger persons roughly half the deaths caused by gallstones are due to gallbladder cancer; for older persons, the proportion of deaths from cancer is smaller because there are more competing causes of death. The absolute risk of cancer is under 1% for all ages and both sexes.

Table 7, column [1], shows that expectant management in general causes a slight loss of life for younger persons, up to 1 or 2 weeks, if no discounting is used. If discounting is used, then expectant management generally causes a modest gain in life expectancy (that is, a negative value for the "loss of life" expectancy).

Prophylactic cholecystectomy does not fare better in these simulations for several reasons. Because elective surgery, or even urgent surgery, which may be incurred by expectant management, is relatively safe, little life is lost by it. Second, if prophylactic cholecystectomy is done and leads to a death at a young age, that death causes a relatively large loss of life expectancy and thus counts heavily against prophylactic cholecystectomy in terms of life expectancy calculations.

Sensitivity Analyses. The results of three sensitivity analyses (columns 2 to 4) assess the importance of several features for which there is uncertainty.

An assumption of a higher rate of cancer, as indicated in columns [3] and [4], would lead to about a 1% cumulative rate of death from gallbladder cancer Table 6 and to a corresponding gain of up to about 2 months in life expectancy for prophylactic cholecystectomy (Table 7).

Increased rates of pain and complications have relatively little effect on results, as shown in column 2. Thus even if the rate of pain is 3.5% per year and of complications is 0.8% per year for 30 year olds, the life expectancy gain of prophylactic cholecystectomy is only a few days longer than in the base case analysis. And if the rate of pain increases to 7%, life expectancy loss from expectant management decreases because of the "protective" effect of receiving a nonurgent cholecystectomy at a young age. Indeed, if the rate of pain increases dramatically, the net result is simply to approach that of "prophylactic cholecystectomy" in which all cholecystectomies are done soon after the discovery of gallstones.

One extreme scenario, shown in column [4], assesses the combined effect of increased rates of pain, complications, and cancer and shows only a modest reduction of life expectancy for expectant management. This scenario demonstrates a paradox that is that increased rates to develop pain or a complication are "protective" overall against the loss of life expectancy by protecting against gallbladder cancer. In other words, when rates of pain and complications are higher (for example, column [4] compared with column [3]), then persons who develop pain or acute complications receive a relatively low-risk cholecystectomy at a younger age, which protects them against the risk for gallbladder cancer later in life. This paradox has been discussed previously [72].

The risk for gallbladder cancer may be an important consideration regarding prophylactic cholecystectomy for some younger persons (for example, age 30 years or under) with asymptomatic gallstones. If the rate of gallbladder cancer was as high as 0.0002 per year for younger persons with gallstones, then prophylactic cholecystectomy would confer an advantage of about 2 months in terms of life expectancy. We believe that the weight of currently available evidence suggests that the lower rates used in the base case are appropriate; however, readers should consider that there is some disagreement about rates of gallbladder cancer, as outlined in section 4.3.1.6 and in Appendix 1. We believe that the data about the rate of gallbladder cancer need to be strengthened, because the data may have important implications for some persons with asymptomatic gallstones.

4.3.1.2. Laparoscopic Cholecystectomy. Laparoscopic cholecystectomy represents a new and very convenient method to treat gallstones, but its safety is yet uncertain. How attractive is this procedure for asymptomatic gallstones? The projections in Table 7 can be used to illustrate the potential benefit of laparoscopic cholecystectomy if it is assumed that the procedure has the same safety as open cholecystectomy. As discussed, some evidence suggests that laparoscopic cholecystectomy may be less safe, at least currently [34], although some persons are optimistic that the procedure will eventually prove to be as safe or safer than open cholecystectomy [74]. The additional convenience of laparoscopic surgery is important and is not reflected in these tables; it must be considered separately in nonquantitative ways.

If laparoscopic cholecystectomy proves to be even safer than open cholecystectomy, then a separate set of calculations can be used to help assess its potential use under those circumstances. The maximum potential benefit of laparoscopic cholecystectomy (or of any therapy for gallstones) is simply the excess risk incurred by a strategy of expectant management compared with the risk of persons who have no gallstones. In other words, the "greatest" gain from any therapy is to reduce the risks of a person with gallstones to the risks of a person with no gallstones. These gains may be estimated by comparing the results of expectant management to those of persons without any gallstones. The gains amount to roughly 1 to 2 weeks more than the gains from prophylactic cholecystectomy.

If life expectancy loss is low, there can be correspondingly little gain from any therapy. In other words, the results suggest overall that asymptomatic gallstone disease generally is not a "serious" problem in terms of mortality (unless gallbladder cancer is thought to be a high risk in asymptomatic persons). The implication is that for any noninvasive or minimally invasive therapy like laparoscopic cholecystectomy to be attractive for asymptomatic gallstones in terms of reducing mortality risk, that therapy would need to be very safe, effective, and convenient.

4.3.1.3. Nonsurgical Therapy. Nonsurgical therapy may be easy and therefore attractive for some persons in whom therapy would be expected to be effective and relatively brief, for example, persons with small cholesterol stones. Nonsurgical therapy does have an added advantage of avoiding surgery. Whether the inconvenience and cost of nonsurgical therapy are worth the effort depends on the magnitude of gain, which is discussed above in the section on laparoscopic cholecystectomy. However, it is uncertain whether nonsurgical therapy would reduce the risk for gallbladder cancer.

4.3.1.4. Gallbladder Cancer. Assessing the potential effect of gallbladder cancer on treatment of asymptomatic gallstones is difficult because the incidence of gallbladder cancer is not well understood. Although the absolute risk of gallbladder cancer is low, gallbladder cancer is almost uniformly fatal, so even a seemingly low rate of cancer of 0.0002 per year would result in approximately a 0.4% risk for death over 20 years and would be an important consideration if young persons with gallstones had rates that high. If there are subgroups of persons with high risks for cancer, then a good case can be made for prophylactic cholecystectomy. It may be that Pima Indians and other New World Indians fall in to such a high-risk group and deserve to be considered for prophylactic cholecystectomy even with asymptomatic stones. Whether some young persons of other races might fall into this category is an important and unresolved question.

4.3.2. Symptomatic Gallstones

4.3.2.1. Open Cholecystectomy. Table 8 and Table 9 show risks of death and life expectancy calculations for persons with symptomatic gallstones who are managed expectantly until a biliary complication occurs. As shown in column [1], for younger persons, expectant management compared with prophylactic cholecystectomy does cause a loss in life expectancy up to about 3 or 4 months (about 1 to 3 months if 5% discounting is used). This gain, although in marked contrast to the results for asymptomatic gallstones, is nevertheless smaller than the several years that may be gained by coronary artery bypass surgery for persons with left main coronary disease [75]. At what point this gain crosses the "therapeutic threshold" is discussed below.

Prophylactic treatment has an advantage in addition to reducing mortality. It also helps avoid an urgent operation which would eventually be needed in about 50% of persons with symptomatic gallstones who might develop a biliary complication like acute cholecystitis. If a patient is averse to the prospect of needing an urgent operation, this benefit may be important in directing the decision about therapy.

4.3.2.2. Laparoscopic Cholecystectomy. In the same manner shown for asymptomatic gallstones, the potential benefit of laparoscopic cholecystectomy may be assessed by considering the maximum gain that any therapy could produce if it had no mortality risk at all. This gain is about 1 to 2 weeks more than the gains shown for prophylactic cholecystectomy.

4.3.2.3. Nonsurgical Therapy. Nonsurgical therapy, such as oral bile acids with or without lithotripsy, may be considered in persons with particularly high operative mortality rates, if the gallstones are amenable to treatment. The gains may be considered as noted in the section above on laparoscopic cholecystectomy. A potential disadvantage of nonsurgical therapy is that it may not reduce the risk for gallbladder cancer.

4.3.2.4. Gallbladder Cancer. Gallbladder cancer appears to be an important consideration in decisions about treatment of persons with symptomatic gallstones because cohort studies suggest that the rate to develop gallbladder cancer appears to be higher in persons with symptomatic stones than in persons with asymptomatic stones. The estimation of the rates of gallbladder cancer is discussed in more detail in Appendix 1. These higher rates are responsible for much of the gain from prophylactic cholecystectomy shown in Table 9.

4.4. Therapeutic Threshold

What is the "threshold" of benefit at which intervention should be chosen for persons with gallstones? The question may be considered as a descriptive problem (that is, "At what threshold do persons select intervention?") or as a prescriptive one (that is, "At what threshold should persons select intervention?") At a descriptive level, perhaps some patients and physicians will "lower" the threshold for intervention for gallstone disease because of the increased convenience of an approach such as laparoscopic cholecystectomy. Assessing whether this is appropriate from a prescriptive standpoint is complicated. In terms of life expectancy, there is relatively little gain from prophylactic therapy for persons with asymptomatic gallstones.

Perhaps some insight into the question of therapeutic threshold may come from a recent analysis that assessed the life expectancy gains from now-adopted interventions that had been studied in clinical trials. The investigators proposed that "life expectancy gains of 3 months or more should be considered to be clinically significant" [76]. By that standard, the prescriptive threshold for intervention for asymptomatic gallstones, at least in terms of mortality, would not be exceeded for any of the natural histories with the exception of natural histories involving very pessimistic assumptions about the rate of gallbladder cancer.

For persons with symptomatic gallstones, there is generally a greater gain in terms of life expectancy, up to 3 or 4 months. When the gain becomes marginal in terms of life expectancy, for example in older persons, the decision will become largely an "elective" one focusing on issues such as the risk for developing pain or the possibility of needing an urgent operation, even if the risk for death is not great. We speculate that for older persons the decision about cholecystectomy may be driven largely by a patient's values and attitudes about the future risk for pain or of needing an urgent operation for possible complications.

An interesting disagreement may ensue between the results of prescriptive analyses (which suggest that there is generally a low mortality from asymptomatic gallstones and most symptomatic gallstones and, at least from the standpoint of mortality, no great urgency to intervene) compared with how patients and physicians will choose to act. Specifically we speculate that patients' values may dominate in decisions about intervention and will be based on personal considerations of convenience and risk for pain.

4.5. Cost and Cost-effectiveness

If the loss of life expectancy is as high as 3 months, then it would take four operations to save 1 year of life, for a direct cost of roughly $32 000 if an operation costs about $8000.

4.6. Comparison with Previous Studies

Other decision analyses and reviews have addressed questions about therapy with cholecystectomy or oral dissolution agents with or without lithotripsy and for asymptomatic and symptomatic stones [18, 72, 77-79]. There is general agreement that asymptomatic stones do not warrant therapy. There is also general agreement that symptomatic stones should be treated, although that strategy has not in the past been uniformly endorsed [72, 80].

One important difference between the "input data" used in this model and in others is the relatively high rates for gallbladder cancer used here, for both asymptomatic and symptomatic persons [72]. The effect of gallbladder cancer, noted in the results section, appears to be potentially important in driving the overall decision about who needs therapy and about the type of therapy, namely whether treatment should remove the gallbladder as well as gallstones. In our view, the data about cancer incidence are potentially important for persons with symptomatic stones.

5. Summary and Discussion of Therapeutic Principles

In the last 15 years, dramatic advances have expanded the options available for treatment of gallstones. As therapies have become safer and easier to use, it has become increasingly attractive to consider treating persons with gallstones. Simultaneously, research has elucidated the prognosis of gallstone disease and has permitted a more selective choice of persons for treatment based on symptom status and projected prognosis.

In making decisions about treatment, clinicians and patients should be guided by considering that management strategies can be aimed at two distinct goals: to prevent biliary pain, or to prevent a biliary complication or death. In all patients with gallstones, whether asymptomatic or symptomatic, the former outcome is much more frequent than the latter. If a patient wishes to avoid any episode of pain, then the natural history data provided in Table 4 and Table 5 may be useful in directing therapeutic decisions. Those natural history data show that persons with asymptomatic gallstones have approximately a 1% to 2% chance per year of having an episode of pain and that the rate appears to decrease over time, whereas persons with symptomatic gallstones have a much higher rate, perhaps as high as 30% to 50% per year for a few years. These simple descriptive data about the risk of pain may, for some persons, be clear and sufficient in guiding a decision about intervention if a patient's primary goal is to avoid an episode of pain. Or it may be helpful to consider life expectancy loss as shown in Table 8 or (Table 9).

We stress that a major clinical challenge in gallstone disease management for primary care physicians is to identify "high-risk" persons. The first task in assessing risk is simply to decide whether gallstones are "symptomatic" because symptom status is so important in predicting all the important outcomes of gallstones: pain, acute biliary complications, and gallbladder cancer. A second challenge is to identify persons among those with asymptomatic or symptomatic stones who have the highest risk to develop an acute biliary complication; there are currently no good predictors.

Important but uncommon risk factors to develop gallbladder cancer include a calcified gallbladder [81] and, perhaps, stones greater than 3 cm in diameter [82, 83].

6. Future Research

Future research should focus on several questions: 1. Natural history of asymptomatic gallstones. Are there subgroups with high risk for developing subsequent pain or complications? Does the rate of complications diminish with time as the rate of pain appears to do?

2. Natural history of symptomatic gallstones. Which persons with symptoms of pain are at highest and lowest risk to develop future episodes of pain or complications? How is subsequent natural history related to initial pain pattern? Are complications regularly preceded by warning symptoms of pain?

3. Gallbladder cancer. Which groups of persons with gallstones, especially groups of young persons, have high risk to develop gallbladder cancer? Is that risk high enough to warrant prophylactic cholecystectomy?

4. Laparoscopic cholecystectomy. How does the safety of laparoscopic cholecystectomy compare with open cholecystectomy? What is the shape of the learning curve for general surgeons who learn the new technique? Which features of patients or surgeons are related to the risk for complications?

Appendix 1. Data Used in the Simulation Model

1. Natural History of Gallstones

For persons with gallstones, rates must be estimated for three outcomes: biliary pain, acute biliary complication, and gallbladder cancer.

a. Persons with asymptomatic gallstones

Biliary pain. The results of several studies suggest that for persons with asymptomatic gallstones the rates of biliary pain are about 1% to 2% per year soon after the detection of gallstones [52, 56, 57] and that the rates decrease over time [52, 56-58]. In the GREPCO study [52], the cumulative rates of symptoms in follow-up were 12.3% at 2 years, 17.3% at 4 years, and 20.5% at 6 years. One possible explanation for the relatively higher rates in earlier years may be that relatively mild pain episodes were included as outcomes. A study by Friedman and colleagues reported an initial rate of 3% to 4% per year, which decreased to 1% to 2% per year during the next 10 years [56].

In the base-case analysis, we use a rate of pain of 0.018 per year for the first 5 years, 0.009 per year for the next 5 years, and 0.0045 per year thereafter. Because of the uncertainty about the rate to develop pain, in sensitivity analysis we assess rates that are as high as 0.07 per year and that do not decrease over time.

Acute biliary complication. Biliary complications appear to be rare as the presenting complaint of gallstone disease. It is not possible, because of the low rate, to estimate directly from cohort studies the rate at which a complication occurs as a presenting biliary problem. At the University of Michigan study of 123 persons followed for 1318 person-years, no person presented with a biliary tract complication [57]. Similarly, in the GREPCO study [52], no person had a biliary complication as the initial biliary problem. In the study by Friedman and colleagues [56], it was not possible to determine whether the complications that occurred were the presenting problems. In six cohort studies that describe initial manifestations, of 91 patients who developed a biliary problem, only 5 had a complication as the initial manifestation [52, 54, 55, 57, 62, 63].

For the rate in the base-case analysis, we use a rate of acute complications of 0.002, which is one ninth the rate of pain for persons at 0 to 4 years after detection of gallstones. This proportion is based on several studies, which suggests that when persons with asymptomatic gallstones have a later biliary problem, a complication is the presenting problem in about 10% of instances [52, 57, 62-64, 86]. In the model, this rate is held constant over time. Because the number of complications reported in cohort studies is so small and because almost no detail is provided about "when" complications occur, it is not possible to determine whether the rate of complications varies over time. In sensitivity analysis we assess complication rates, which are 0.008, or four times as high as in the base-case analysis.

Gallbladder cancer. Two approaches are used to estimate rates of gallbladder cancer. In a direct approach, cohort data are used to assess rates directly. A second, indirect approach is to estimate rates by obtaining "numerator" data as the number of cases of gallbladder cancer and the "denominator" data as persons with gallstone disease and to estimate the rate of cancer among persons with gallstones.

Direct approach. In asymptomatic persons, eight cohort studies [52, 54, 56, 57, 59, 62, 63] provide a total of 854 persons followed for about 6339 person-years of follow-up; no gallbladder cancers have been reported in this group.

It is possible to assess statistically whether the rate obtained in the cohort studies is likely to be obtained by chance if the true rate is as high as 0.0006, which is the rate used in the upper end of the sensitivity analysis. Using the binomial probability distribution, the probability of observing 0 cancers in 6339 person-years of observation is P < 0.05 when the true rate is 0.0006. Indirect approach. This method [72, 87] uses noncohort data to estimate cancer rates for persons of different ages. For numerator events, age-and sex-specific numbers of deaths from gallbladder cancer are used. Denominators, consisting of the persons at risk, are estimated from age-specific prevalence of gallstones and the population of the United States.

In the model, the rates we used in the base case for persons under age 50 years were derived from the indirect method described above and are shown in Appendix Table 1. For persons ages 50 years and older, the rate used was 0.0002 per year, a rate derived from cohort data of persons with symptomatic stones, which is one quarter the rate derived from those cohort studies. The reasoning is as follows: About 50% of persons with gallbladder cancer have had symptoms [88]. Of persons with gallstones, about 10% to 30% are symptomatic [52, 62, 89]; the rate seems to vary, in part, with the definition and criteria for symptoms. Because about half the cancers occur in persons with symptoms and there are about one quarter as many persons with symptoms as without, then the rate of cancer in persons with symptoms is about four times as high as in persons without symptoms. In sensitivity analysis we assess gallbladder cancer rates as high as 0.0006 per year, which would be equivalent to a cumulative risk of about 0.012 over 20 years. Such a rate is, in our view, above the bounds of the true rate that could be reasonably expected for persons with asymptomatic gallstones based on the results of the cohort data described above.

b. Persons with symptomatic gallstones

Biliary pain. For persons with symptomatic gallstones (those that have caused pain), the subsequent natural history is "ominous" in that recurrent episodes of pain are quite common. Some evidence suggests, however, that about 30% of persons with one pain episode may have no further episodes [58, 62, 64, 68]. In the model, we do not consider the outcome of biliary pain explicitly; in the model for symptomatic stones it is assumed that persons have decided to try to "risk" future episodes of pain in the hopes of being among the 30% of persons who do not have recurrent pain and that they may not need an operation for a complication. The rates of recurrent pain are so well documented and so high that a person who wants to avoid all episodes of pain should simply choose prophylactic therapy.

Biliary complication. To estimate the rate that a person with biliary pain develops a biliary complication, data from several cohort studies were used. Because so few results were presented actuarially in life-table form, we calculated actuarial results where possible using the data presented by the authors. Newman and colleagues reported for 362 persons who had had biliary pain and then follow-up averaging 10 years that "during each year after the first, one of the five major complications occurred in 1.5% ..." [64]. In the National Cooperative Gallstone Study [7], of 112 persons with pain during the year before entry into the study, 12 persons had cholecystectomy that was "clinically indicated for biliary pain" [85]. Four complications (three acute cholecystitis and one pancreatitis) occurred in approximately 513 person-years of follow-up, for a rate of 0.78% per year. McSherry and colleagues [59] reported 15 complications in about 1539 person-years of follow-up (1.2% per year). The GREPCO study reports two biliary complications in 106 person-years of follow-up (0.9% per year) [62]. Friedman and colleagues report a rate of about 1% per year, which appears to be constant over 20 years [56]. Based on these results, we used for the base case a rate of 1.5% per year in the model and assessed insensitivity analysis rates as high as 5% per year.

Gallbladder cancer. The direct approach was used to estimate rates of gallbladder cancer for symptomatic persons. Eleven cohort studies [7, 56, 58, 59, 61, 62, 64, 68, 90-92] report on 4781 persons with about 32 134 person-years of follow-up. Twenty-five cancers occurred, for a rate roughly of 0.00078 per year.

In the base-case analysis, we use a rate of 0.0008 per year for symptomatic persons older than 50 years (Appendix Table 2). For persons younger than 50 years, we use a rate four times as high as that derived using the indirect approach for persons with asymptomatic stones, as described above. In sensitivity analysis we use rates twice as high, or 0.0016 per year. The importance of assumptions about the rate of gallbladder cancer is discussed in the text.

2. Operative Mortality Rates for Cholecystectomy See (Table 2).

3. Mortality Rates from Other Causes See reference [72].