The Choice of Primary Repair or Mesh Repair for Paraesophageal Hernia

A Decision Analysis Based on Utility Scores

Nadia M. Obeid, MD; Vic Velanovich, MD


Annals of Surgery. 2013;257(4):655-664. 

In This Article


Decision analysis begins with construction of a "decision tree" starting from a choice.[18] The choice, in this case, is whether or not to perform a PEH repair with or without mesh. For the purposes of this analysis, the choice will be limited to primary repair versus mesh repair. Other considerations, such as need for a Collis esophageal lengthening procedure, use of gastropexy or fundoplication, minimally invasive versus open repair, etc, will not be addressed here, and therefore, will have no bearing on the choice. For the purposes of this study, "primary repair" is defined as a suture repair to approximate the left and right diaphragmatic crura; a "mesh repair" is defined as any type of repair, using any type of mesh, in any configuration. There will be no attempt to assess the "best" type of mesh repair. Decision analysis requires a literature search to obtain information as to the possible consequences of the decision and the likelihood of their occurrence.

Data Sources

A literature search through using the phrase "paraesophageal hernia repair" was completed. The subsequent articles that meet the following inclusion criteria were further reviewed: patients had to have had some type of PEH repair (mesh repairs of type I hiatal hernias were excluded); and information from the article had to include, in addition to the type of repair, at least 1 of the following: enough operative detail to properly categorize the repair, perioperative complications, length of follow-up, recurrence rate, reoperation rate, symptomatic recurrence rate, and outcome of reoperation. Table 1 summarizes the definition of terms used in this decision analysis. The initial literature review was to include all available literature from the beginning of the laparoscopic PEH repair era from the early 1990s. However, 4 systematic literature reviews and meta-analyses were published in 2006 and 2007, which synthesized the available literature up to that point.[19–22] Therefore, these articles were used as data sources for the era of 1990 to 2006. The reference lists from these articles were then reviewed for potentially additional sources of data. During the review process, many centers published more than 1 study of their PEH repair results. To avoid overweighting of data from centers with multiple publications, when more than 1 study was published by a center, the latest published study with the largest sample size and the longest follow-up was used unless the center published other studies which contained information that was useful for the decision analysis in one study but not in another. Table 2 [19–45] and Table 3 [21,26–31,33–38,41,42,44–49] present the references with the data sources used in this decision analysis.

Construction of the Decision Tree

The decision tree is constructed with choice nodes (points at which the decision-maker must make a choice) and chance nodes (points at which different outcomes can occur depending on the chance of those outcomes occurring).[50] Conventionally, a decision tree is written and read from left to right. In this decision analysis, only 1 choice will be analyzed, namely the choice of performing a primary suture or mesh repair of an initial, elective PEH. There will be no differentiation in the type of mesh used or the technique of placement. Although it is acknowledged that there is controversy as to the mesh choice and outcomes may depend on the mesh choice, these potential differences will be addressed in the sensitivity analysis described later.

After the initial choice of the repair type, 2 outcomes can occur: either the patient recovers without a complication, or the patient suffers some type of short-term complication. A short-term complication will be defined as either perioperative death, or some other complication from which the patient will eventually recover, such as a wound infection, pulmonary, cardiac, or some other complication. Only those complications that could be classified as Clavien class II, III, or IV (ie, those requiring some type of intervention)[51] will be acknowledged in this decision analysis. The next chance occurrence is that of PEH recurrence: either the patient will have a recurrence or will not. The literature acknowledges that some recurrence will be symptomatic and some asymptomatic; however, it is not always clear whether symptoms were the driving force behind the decision to reoperate. Therefore, the next chance node will be reoperation. Although it is acknowledged that the decision to reoperate or not is a choice, for the purposes of this decision analysis, it will be treated as a chance, using the reported frequency of reoperations. Of the patients with recurrence who did not undergo reoperation, their recurrences were either symptomatic (and presumably treated conservatively) or asymptomatic (and therefore were considered not to affect their overall health and well being). Of the patients undergoing reoperation, they either had a good outcome, a severe adverse outcome, or perioperative death (see Table 1 for definitions).

Standardization of Outcome/Utility Score

Many decision analyses will base their outcome analysis on the concept of quality-adjusted life expectancy,[18,50] which combines the effects of the treatment on the patient's life expectancy and the patient's quality of life. This is usually done by quantifying quality of life using a "quality multiplier" that ranges from 0 (death) to 1 (perfect health) and multiplying this fraction and the life expectancy. As PEH is a benign disease that rarely (albeit dramatically) causes death as part of its natural history, and can occur at any age, we believe that the effect of operative treatment would be best standardized using a utility score. To understand the utility score, we need to understand how to quantify patient "preferences." A utility score assesses an outcome based on a "preferred" health state for the patient. The scale is defined from the "most preferable" state being 1 to the least preferable state being 0; other states are then fractions between 0 and 1.[18] Typically and intuitively, a score of 0 is death and a score of 100 is perfect health.[52,53] An outcome that has a score of 0.5 is defined as the state at which a patient will take a 50/50 chance on the best outcome or the worst outcome occurring. For this study, the least preferred outcome is perioperative death, and it will be assigned a score of 0. The most preferred outcome is a successful PEH repair, with no recurrence, no adverse outcomes, and resolution of preoperative PEH symptoms (the "perfect" outcome). We are then left with assigning utility scores for the other possible outcomes—namely, a patient who has some type of short-term complication, from which there will be a complete recovery; a recurrence that is not reoperated upon and is asymptomatic; a recurrence that is not reoperated upon and is symptomatic; a patient who is reoperated upon and has a good outcome; and a patient who is reoperated upon and has a severe adverse outcome (esophagectomy and/or gastrectomy, stricture, or erosion). Assigning utility scores can be accomplished in several ways. One way is to ask patients to assign scores after discussions as to the implications that each state would have on their lives, another is by expert proxy, and the third is to review the available literature on the clinical consequences for the patient.[18,52,53] We have previously demonstrated in pancreatectomy patients that patient-derived utility scores correlate well with patient-derived quality of life scores using a validated, generic quality of life instrument.[54] Therefore, we will use this method in this study.

For the purposes of this study, the prima facie rank order of the possible outcome states would be the perfect outcome > good outcome with short-term perioperative complications > reoperation with good outcome > symptomatic recurrence > severe adverse outcome (ie, esophagectomy, gastrectomy, erosions, or strictures) > death. We have discovered that the outcome state of a patient who undergoes a PEH without complications or recurrence with a good outcome will have a utility score of 100, and the outcome state of a patient suffering perioperative death will have a score of 0. The outcome state of a patient who has some type of complication will be given a score of 98 (it is clearly preferable not to have a complication than to have one, but it is preferable to have a short-term complication that will result in a good outcome than a recurrence or other adverse outcome). The outcome state of a patient who has a recurrence but is not reoperated upon and is asymptomatic will be given a score of 100 (based on the premise that from the patient's perspective, a conversion of a symptomatic PEH to an asymptomatic hiatal hernia would be the same as not suffering from a PEH at all), whereas one who is symptomatic is given a score of 90, based on existing quality of life data of patients with symptomatic PEH, which shows that quality of life scores for patients with PEH are about 10% lower than the general population.[2,37,44,55] The outcome state of a patient who has a reoperation and has a good outcome will have a score of 95 (it is clearly preferable not to have a reoperation, but a reoperation with a good outcome is preferable to continued symptoms. Therefore, the score was assigned midway between the most preferred state and the symptomatic state), whereas a patient who has a reoperation and suffers a severe adverse outcome will have a score of 80, based on the existing quality of life data for patients after esophagectomy.[56–58]

Figures 1 and 2 show the decision trees for primary repair and mesh repair using the average or predominate frequencies of the occurrences as determined from the available literature (Table 2 and Table 3).

Figure 1.

Decision tree for primary repair. The fraction underneath the branch is the frequency of that particular outcome occurring. At the terminal of the branch is the utility score for that particular outcome.

Figure 2.

Decision tree for mesh repair. The fraction underneath the branch is the frequency of that particular outcome occurring. At the terminal of the branch is the utility score for that particular outcome.

Sensitivity Analysis

It is clear from Table 2 and Table 3 that there is significant variation in the reported rates of the important occurrences after PEH repair for either primary or mesh repair. The main points of variation are the recurrence rates of PEH for primary and mesh repair and the need for reoperation if a recurrence occurs. Perioperative complications and death are quite similar between the 2 types of repair; therefore, sensitivity analysis is not needed for this aspect of the decision analysis. Similarly, the frequency of symptomatic recurrences and severe adverse outcomes is more dependent on the rates of recurrence and the need for reoperation; therefore, sensitivity analysis of these aspects of the decision would not greatly affect the utility score as much as recurrence and reoperation. However, the combination of the recurrence rate and reoperation rate is important in determining the utility score, so this combination will be analyzed by sensitivity analysis. Because we wished to analyze the effects of recurrence and reoperation rates, we will exclude perioperative complications from this sensitivity analysis, as they are similar between both types of repair and would not greatly influence the utility scores.