Appendicitis, Adult

Prior Probability

The incidence of appendicitis among emergency patients with abdominal pain is up to 25% for patients younger than 60 years. For those older than 60 years, the incidence is up to 5%.

Populations for Whom Appendicitis Should be Considered

• All patients with abdominal pain.

Detecting the Likelihood of Appendicitis Among Emergency Patients With Abdominal Pain in the RLQ

The Alvarado¹ model is recommended as the most user-friendly while being among the most powerful. The details of the clinical decision rule are displayed in Table 5-6. Note that the MANTRELS mnemonic is helpful in that it is easy to remember and is organized according to medical history, physical examination, and laboratory data.

Reference Standard

Histologically proven diagnosis or no surgery after adequate follow-up (which allows the inference that appendicitis was not present).

Clinical Scenario

A 29-year-old patient presents to your office with abdominal pain and a fever. The patient was well until 1 day ago and had never experienced abdominal pain. A vague periumbilical pain awoke him from sleep 12 hours previously, and he soon developed anorexia, nausea, and vomiting. His wife consulted their family medical reference guide and then brought him to the office, concerned that his symptoms matched a description of appendicitis. The pain then migrated to the right lower quadrant (RLQ) and was much worse while he was riding in the car to the physician's office.

The patient's oral temperature is 37.8°C; the pulse rate and blood pressure are normal. He has RLQ tenderness, guarding but not rigidity, and rebound tenderness in the RLQ. A rectal examination reveals no tenderness, and he does not exhibit the psoas or obturator signs. Rovsing sign is positive.

Why Is This an Important Question to Answer With a Clinical Examination?

In western countries, appendicitis represents a common cause of acute abdominal pain. According to National Center for Health Statistics data, approximately 500 000 patients underwent appendectomies from 1979 to 1984. Individuals carry a 7% lifetime risk of developing appendicitis.¹ The incidence of appendicitis causing abdominal pain depends on the clinical setting. In series from emergency departments or surgical services, 25% of patients younger than 60 years and evaluated for acute abdominal pain have acute appendicitis, whereas the incidence in those older than 60 years is approximately 4%.^1-5 Only 0.7% to 1.6% of all ambulatory patients with abdominal pain have appendicitis.^{6, 7} Among children treated in the ambulatory care setting, appendicitis causes 2.3% of all abdominal pain episodes.⁸ In children admitted for acute abdominal pain, appendicitis is the etiology for approximately 32%.^9-11

The morbidity and mortality of appendicitis remain significant, even with the advent of antibiotics and effective surgical management. Although the overall mortality rate with appropriate treatment is less than 1%, in the elderly it remains approximately 5% to 15%.^{2, 4} There is a significant amount of morbidity caused by appendiceal rupture.^12-15 The incidence of perforation in patients with appendicitis ranges from 17% to 40%, with a median of 20%.^{16, 17} The perforation rate is significantly higher in the elderly, with rates as high as 60% to 70%. Several factors contribute to the increased incidence of perforation in the elderly, including significant delay in seeking care, nonspecificity of the presenting symptoms and signs, diminished febrile response, and fewer abnormalities in important laboratory characteristics such as the white blood cell count (WBC).^{2, 3, 5, 14, 18, 19} Children also have an increased incidence of perforation because of delays in consulting a physician for abdominal pain.⁸ The negative laparotomy result rate in most series ranges from 15% to 35% and creates morbidity.^{16, 17, 20-22} In younger women, the negative laparotomy result rate is significantly higher (up to 45%) because of the prevalence of pelvic inflammatory disease and other common obstetric and gynecologic disorders.^{16, 17, 23, 24}

The Accuracy of Other Diagnostic Modalities

Routine medical history and physical examination remain the most effective and practical diagnostic modalities.^{25, 26} Several other clinical methods for diagnosing appendicitis have been studied. Computer or algorithm-driven analyses of patients with abdominal pain have been evaluated,^27-35 although most studies have incomplete controls and yield inconsistent results. Thus, the utility of computer-guided diagnosis compared with unassisted clinical diagnosis needs further evaluation. The authors of most of these studies believe that the improved utility they demonstrated was primarily because clinicians were forced to focus on specific clinical data that were readily available to be entered into the analysis tree. Finally, these authors observed that all of these modalities completely depend on the accuracy of the data gathered and interpreted by clinicians before the data are entered into the computer or algorithm analysis. The concept of an extended period of observation of patients with questionable appendicitis has been shown by some authors to be helpful.^{8, 27, 28} Its utility, like that of computer and algorithm analyses, depends on routine medical history and physical examination skills of clinicians.

The utility of radiographic techniques has also been evaluated. Plain abdominal radiographs and barium enemas are neither specific nor sensitive for appendicitis.³⁶ Ultrasonography is more effective in detecting a distended appendix than appendiceal perforation.^{10, 15, 36-44} No study has demonstrated ultrasonography to be clearly superior to the clinical examination, and many authors believe that its primary utility is to supplement the medical history and physical examination in patients with equivocal findings. The accuracy of computed tomography in diagnosing appendicitis has also been inconsistent.^{36, 42, 43}

Laparoscopy has been shown by some authors to be useful, particularly in young women in whom it can be difficult to differentiate between pelvic inflammatory disease, ectopic pregnancy, and appendicitis.²⁷ However, other series have not been as supportive, with negative appendectomy result rates from 20% to 30%.^{44, 45} Studies of outcomes comparing laparoscopy with laparotomy have yielded conflicting results.^{46, 47} Even though ultrasonography, computed tomography, and laparoscopy can be helpful, none are ideal techniques, and the clinician must depend on patient medical history and physical examination results.

Appendiceal Anatomy and Pathophysiology of Appendicitis

The adult's appendix averages 10 cm in length, arising from the posteromedial wall of the cecum, about 3 cm below the ileocecal valve.⁴⁸ Its position in the abdominal cavity is variable, being described as retrocecal, retroileal, preileal, subcecal, or pelvic, and this variability in location may influence the clinical signs and symptoms associated with appendicitis. Although the physiologic role of the appendix is unproved, an immunologic function is suggested by its content of lymphoid tissue.⁴⁹

Appendiceal obstruction, followed by secondary bacterial invasion, causes the majority of appendicitis. Continued fluid secretion by the mucosa of the obstructed appendix distends the lumen, eventually exceeding venous pressure and leading to tissue ischemia and, ultimately, necrosis. Causes of obstruction include fecaliths, calculi, tumors, parasites, foreign bodies, or, rarely, barium. In the one-third of patients without apparent obstruction, infection by viruses, parasites, or bacteria, or either trauma or postoperative fecal stasis may be involved.^50-55

Normally, appendicitis presents with a highly characteristic sequence of symptoms and signs.⁵⁶ Initially, appendicitis causes visceral pain poorly localized to the epigastrium or periumbilical region, presumably because of distention of the appendix. Anorexia, nausea, and vomiting soon follow as this pathophysiology worsens. More advanced inflammation causes irritation of adjacent structures or the peritoneum, low-grade fever, and peritoneal pain localized to the RLQ. The pathophysiology explains the classic migration of pain caused by appendicitis. The point of maximal tenderness may be distinct from McBurney point, 5 cm from the anterior superior iliac spine on a line running from the umbilicus.

Atypical locations of the appendix may lead to unusual clinical findings. In the case of retrocecal or retroiliac appendices,^{57, 58} the pain may be poorly localized and may not undergo the transition from epigastric to RLQ locations. Pelvic appendicitis frequently causes pain in the left lower quadrant, with an absence of tenderness, and is reflected by increased pain during a rectal examination. Unusual symptoms of urinary and defecation urgency, caused by irritation of the ureter and rectum, respectively, plus dysuria and diarrhea may also occur.

Although often a diagnostic dilemma in the first trimester of pregnancy because of confusion with other diagnoses, appendicitis in later stages of gestation may present a challenge for the clinician because of displacement of the appendix by the enlarging uterus. In such cases, periumbilical or right subcostal tenderness may be found.

How to Elicit the Relevant Symptoms and Signs

Pain is commonly the first symptom of appendicitis.^{9, 59} Classically, the vague, midepigastric or periumbilical pain awakens the patient from sleep but is not initially severe. After reaching its peak in around 4 hours, it diminishes and then migrates to the RLQ. Most patients will seek medical attention within 12 to 48 hours. Pain usually occurs before vomiting, and the patient has usually not experienced similar symptoms before the present episode.

According to Cope's Early Diagnosis of the Acute Abdomen,⁶⁰ many patients feel constipated and anticipate that defecation will relieve discomfort, leading them to use cathartic agents. However, pain persists after a bowel movement.

Many signs have been associated with appendicitis or peritonitis. Some of obvious value, such as the pelvic examination, have not been adequately evaluated to merit mention in this systematic review or they lack an adequate description or standardization of the elicitation of the sign to ensure accurate reproduction. A common reference for definitions in the best studies is a text by De Dombal.⁶¹ What follows is the most consistent and useful description of the signs:

• Guarding: Guarding is a state of voluntary contraction of the abdominal muscles. The muscles are held tense by the patient because he or she knows (or fears) that further examination is likely to be painful. Fear can be partially, or fully, overcome by tact and persuasion.⁶¹

• Rigidity: Rigidity is also known as involuntary guarding. The best studies of abdominal pain have described rigidity as an involuntary reflex spasm of the muscles of the abdominal wall. It can never be overcome by tact and reassurance.⁶¹

• Rebound tenderness: (1) Press on the area of question with the flat of your hand, sufficient to depress the peritoneum. The patient should be experiencing pain. (2) Keep pressing with a constant intensity. As the patient adjusts to this pressure during 30 to 60 seconds, the pain diminishes. It may go away completely, although usually it does not. (3) Without warning, and preferably while the patient's attention is distracted, remove the hand suddenly to just above skin level. Watching the patient grimace is more indicative than a complaint of pain.⁶¹

• Rovsing sign: A sign related to the rebound tenderness test. Press deeply and evenly in the left lower quadrant and then release pressure suddenly. The presence of tenderness in the RLQ during palpation or referred rebound tenderness in the RLQ during release is considered a positive Rovsing sign.

• Psoas sign: With the patient in the supine position, ask the patient to lift the thigh against your hand, placed just above the knee. Alternatively, with the patient in the left lateral decubitus position (Figure 5-1), extend the patient's right leg at the hip. Increased pain with either maneuver is a positive sign and indicates irritation of the psoas muscle by an inflamed appendix.

• Obturator sign: This sign is similar mechanically to the psoas sign. It is elicited by passively flexing the right hip and knee and internally rotating the leg at the hip, stretching the obturator muscle (Figure 5-2). Resultant right-sided abdominal pain is a positive sign, indicating irritation of the obturator muscle. The obturator sign has not been studied independent of the psoas sign, but most clinicians would attribute the same significance.

• Rectal examination: Classically, tenderness and fullness perceived on the right but not the left side on rectal examination are indicative of a pelvic appendicitis.⁶⁰ This sign is subjective and poorly described in most major physical examination texts. No studies that assess rectal tenderness describe the examination technique.

Precision of These Symptoms and Signs

There have been no studies published evaluating the precision of the clinical examination for appendicitis. A standardized clinical examination might produce strong interrater reliability.

Accuracy of These Symptoms and Signs

A handful of studies published during the past few decades have evaluated the accuracy of the clinical presentation of appendicitis. The studies are of various quality and design. Most are best described as cross-sectional in design because a clinical judgment is made, with outcomes measured in terms of pathologic confirmation of appendicitis vs a negative laparotomy result or no requirement for surgery. Eleven of the highest-quality studies, based on number of patients studied, the study design, and completeness of reported data, are summarized in Table 5-1.^{9, 24, 33, 35, 62-67} The search strategy for identifying these articles is available from the authors on request. This strategy yielded about 300 articles since 1966. Further limiting sets to adult age groups yielded 200 studies. The titles and abstracts were reviewed and chosen if adequate detail of the outcomes and aspects of the clinical examination allowed construction of 2 × 2 tables and subsequent calculation of likelihood ratios [LRs].

The 11 studies were divided into 2 groups by the patients on whom they focused. Approximately half of the studies focused on patients in whom appendicitis was suspected, and half, on those who were examined for acute abdomen. In the studies of suspected appendicitis, the inclusion criteria were not further defined. In the studies of acute abdomen, inclusion criteria usually involved pain for less than 1 week. Taken together, the studies report on the findings of more than 4000 patients and provide the best available evidence supporting the most valuable aspects of the clinical examination for appendicitis (Table 5-2).

Each study reports on a varying constellation of clinical findings. Many aspects of the clinical examination are not evaluated in all of these studies. Unfortunately, some of the aspects evaluated are poorly defined in the text of the studies, so specific recommendations for these aspects are difficult to derive for medical education or the everyday practice of medicine.

Nonetheless, several points can be drawn from a systematic literature review. In evaluation of patients presenting with emergency and acute abdominal pain, usually defined as less than 1 week in duration before presenting to an emergency department or surgical ward, the prevalence (pretest probability) of acute appendicitis ranges from 12% to 26%.^{12, 30, 32, 69} The clinical examination will influence this probability further. If various aspects of the clinical examination are viewed as diagnostic tests, LRs ^{70, 71} and posttest probability can be calculated.

From the medical history, 6 aspects have been evaluated. Seven physical examination items have also been studied well. These aspects are examined further in Table 5-3.⁷² The large number of patients studied and the similarities across studies make the data suitable for being combined into summary measures.

Three findings show a high positive LR (LR+) across all studies and, when present, are most useful for identifying patients at increased likelihood for appendicitis: RLQ pain (LR+, 8.0), rigidity (LR+, 4.0), and migration of initial periumbilical pain to the RLQ (LR+, 3.2). Rebound tenderness was studied in most patients, but its positive likelihood varied too much to allow a statistical point estimate of its effect (LR+, 1.1-6.3). Although the obturator sign has not been studied independently, the authors suspect that this sign has operating characteristics similar to those of the psoas sign.

Clinicians also collect evidence to help prove normality. Unfortunately, no single component consistently provided a low negative LR (LR–) that would rule out appendicitis. There were, however, many signs that proved to be helpful in ruling out appendicitis. The absence of RLQ pain and the presence of similar previous pain demonstrated powerful LR– (0.28 and 0.25, respectively). The absence of the classic migration of pain also diminished the likelihood of appendicitis significantly (LR–, 0.5). The absence of RLQ guarding or rebound pain has excellent properties for ruling out appendicitis in some studies, but not others. The presence of pain before vomiting needs further study to identify its diagnostic efficiency because, in its only evaluation, it was highly efficient in ruling out appendicitis. Astute clinicians will recognize that the absence of anorexia, nausea, or vomiting has little effect on the likelihood of appendicitis.

The Role of Combined Findings

Clinicians rarely rely on a single sign or symptom for diagnosis but instead rely on a combination of findings. Unfortunately, the precision and accuracy of combinations of findings have not been reported in these studies. Several studies do assess, however, various decision rules that do combine these findings.^{6, 33-35, 66, 73-77} Four of the most powerful rules were validated on an independent set of 1254 patients older than 50 years and presenting with abdominal pain. No single score was found to be superior; however, it was observed that the decision rules reported in the original work to be most powerful incorporated at least 2 of 5 common variables: site and duration of pain, site of tenderness, rebound tenderness, and leukocytosis.⁷⁸

The Bottom Line

Returning to the beginning clinical scenario, the historical components of the presentation are highly suggestive of appendicitis. Our patient demonstrates the classic sequence of abdominal pain before vomiting, culminating with the migration of the initial midepigastric pain to the RLQ. The combination of these LR+s alone makes appendicitis more likely.

The findings of guarding but not rigidity tend to neutralize each other's effect. The rectal examination results and the psoas and related signs are helpful if present but are not helpful when absent, as in this case. In sum, we suspect appendicitis in this man, so further evaluation is warranted.

A surgical doctrine suggests that a decrease in the perforation rate will be achieved only by an increase in the negative laparotomy result rate in suspected acute appendicitis. The truth of this doctrine has been called into question, given the results of large- and small-area variation studies.²⁹ Improved clinical evaluation is suggested as a remedy for a high rate of negative laparotomy results without increasing the perforation rate. Evidence suggests the essential nature of clinical details.^{79, 80} Clinicians often do not collect enough clinical details for accurate and precise diagnosis.^81-83 Correction of this deficit, therefore, may well increase diagnostic accuracy without increasing the perforation rate.

In summary, there are several conclusions that can be made concerning the clinical presentation, pathophysiology, and diagnosis of appendicitis:

1. Appendicitis is a common clinical entity, with significant morbidity and mortality, particularly at the extremes of age.

2. The pathophysiology of appendicitis consists of initial dilatation of the appendix, followed by appendiceal ischemia, necrosis, and parietal peritoneal irritation. Clinical findings are predictable, predicated on knowledge of this pathophysiology.

3. The characteristic sequence of symptoms and signs includes the following: (1) vague pain initially located in the epigastric or periumbilical region; (2) anorexia, nausea, or unsustained vomiting; (3) migration of the initial pain to the RLQ; and (4) low-grade fever.

4. Migration of pain in the characteristic manner, RLQ pain, and the presence of pain before vomiting are historical findings that suggest appendicitis. The presence of rigidity, a positive psoas sign, fever, or rebound tenderness is a sign on physical examination indicating an increased likelihood of appendicitis.

5. Conversely, the absence of RLQ pain, the absence of the classic migration of pain, and the presence of similar pain previously are powerful symptoms in the medical history that make appendicitis less likely. In the physical examination, the lack of RLQ pain, rigidity, or guarding makes appendicitis less likely.

6. Because no finding on the clinical examination can effectively rule out appendicitis, prudence dictates close follow-up of patients with abdominal pain who do not receive further diagnostic testing.

Author Affiliations at the Time of the Original Publication

Department of Internal Medicine, University of Texas, Southwestern Medical Center, Dallas, Texas (Drs Wagner, McKinney, and Carpenter).

Acknowledgments

We appreciate the expert advice offered by gynecologist Joanne Piscitelli, MD, and general surgeon Ted Pappas, MD, both of Duke University, Durham, North Carolina, during the preparation of the manuscript.

References

Clinical Scenario

A 24-year-old woman presents with abdominal pain, nausea, and vomiting. She describes the pain as beginning in her midabdomen 3 days ago, and it has gotten progressively worse. Her last menstrual period was 3 weeks ago and was normal; she is not sexually active. The pain has stayed in the midabdomen and not moved to other locations. On examination, she has a fever and right lower quadrant (RLQ) and rebound tenderness; her pelvic and rectal examination results are unremarkable. Laboratory evaluation reveals a left shift without leukocytosis and ketonuria.

Updated Summary on Adult Appendicitis

Original Review

Updated Literature Search

Our literature search used the parent search for the Rational Clinical Examination series, combined with the subject headings “exp appendicitis” published between 1994 and September 2004. This search yielded more than 400 titles, which were narrowed down to approximately 50 by excluding studies of laboratory and radiologic tests and case studies.

There have been few new studies that focused on the operating characteristics of individual components of the clinical examination for appendicitis. However, there have been several studies that have looked at combinations of findings. That is, instead of examining the likelihood ratio (LR) of rebound tenderness alone, studies have explored the combination of fever, migration of pain, and rebound tenderness.

The studies of clinical decision rules were selected if the components, derivation, and validation of the prediction rule were clearly defined in the article and the patients included were those from a general population with abdominal pain or were suspected of having appendicitis. Our previous literature search was reviewed, and studies conducted before 1994 were included if they fit these criteria.

New Findings

• Combinations of findings from the clinical examination are more powerful than any single finding.

• Most of the decision rules formed by these combinations of findings include migration of pain from periumbilical to RLQ, rebound tenderness, RLQ tenderness, nausea-vomiting, male sex, fever, rigidity, and white blood cell (WBC) count.

Details of the Update

Eighteen studies that derived or validated clinical decision rules for appendicitis were identified. The most important studies were those by Alvarado,¹ Eskelinen et al,² and Fenyo et al.³ These studies were chosen because of their methodology, large sample sizes, simplicity of the decision rule, or familiarity with physicians. In addition, a study that compared several clinical decision rules on the same population provided a good perspective of the relative value of these rules.⁴

The Alvarado¹ study was one of the first of the clinical decision rules published, demonstrating the power of the rule beyond individual findings. Although the methods are rudimentary and the rule is not validated in the study, it represents the most widely accepted and the simplest of the clinical decision rules. By combining the results for 8 findings from the medical history or the examination (which conveniently spells out the mnemonic MANTRELS), the resulting score provides guidance on whether to operate in the setting of suspected appendicitis. Of 10 potential points, patients with a score of 7 or higher are recommended for surgical intervention. The various components are Migration, Anorexia-acetone, Nausea-vomiting, Tenderness in RLQ, Rebound pain, Elevation of temperature, Leukocytosis, and Shift to the left of normal WBC count.

The Eskelinen et al² study evaluated more than a thousand patients with a rule that includes 7 variables in men and 5 in women. The disadvantages of this study are that the rule is complex, computer based, and was validated with a small number of patients.

The Fenyo et al³ study assessed 10 variables used in a complex equation. The results for the individual findings showed that a WBC count of less than 8.9 × 10⁹/L (LR, 0.16) was the one finding that had reasonable measurement properties, leading to a lower likelihood of appendicitis.

The Ohmann et al⁴ study displayed a parallel analysis of 10 available studies, including the 3 mentioned above. A database of 45 variables prospectively collected from 1254 consecutive patients on a standardized form was used to evaluate these studies. A surprising outcome of the study was that none of the rules produced sufficiently low rates (<15%) for either unneeded appendectomy result (rule advised surgery but normal appendix found) or delayed appendectomy (rule advised delay but the patient proved to have appendicitis). However, the clinicians in these studies did not perform much better than the rules. Although the clinicians who chose not to use the rules performed similarly to the decision rule results, implementing decision models in actual clinical practice may identify a subset of clinicians who improve with the rules. The authors recommend the Alvarado¹ and Eskelinen et al² studies as those warranting further evaluation.

What lessons can be learned from these studies? The rules recommended by experts incorporate a description of the pain location (and change of location) from the medical history, rebound and RLQ tenderness on the physical examination, and leukocytosis. Other commonly included variables are nausea, vomiting, male sex, fever, and rigidity. Decision rules do not vary dramatically between women and men with abdominal pain.

Improvements in the Data Presented in the Original Publication

A series of letters to the editor prompted by the original publication pointed out some ways the presentation of the data could be improved.^{5, 6} The numbers reported for the sensitivities and specificities did not match the reported LRs; this error was explained⁷ and corrected in Table 5-3 of the original publication.

Changes in the Reference Standard

There were no changes in the reference standard; appendicitis is still a histologically proven diagnosis, and the absence of appendicitis is still a clinical diagnosis (ie, no surgery after adequate follow-up). A recent systematic review suggests that computed tomography may be more accurate than ultrasonography for identifying patients with appendicitis, but neither test is sufficient to serve as the reference standard.⁸

Results of Literature Review

The LR and diagnostic odds ratio of the 8 best clinical decision rules are displayed in Table 5-4. The studies with the highest numbers of participants that evaluated rules with the highest diagnostic odds ratios (a measure of overall accuracy) were Alvarado,¹ Fenyo et al,³ and Eskelinen et al.²

Although the approaches of Fenyo et al³ and Eskelinen et al² have a higher diagnostic odds ratio than the Alvarado¹ rule, both have a large number of variables and require multivariate modeling that make them hard to use without a handheld calculator or coding sheet. According to these results, as well as expert opinion expressed by the parallel evaluation of Ohmann et al⁴ of 10 of the studies, the Alvarado¹ clinical prediction rule (Table 5-5) is used by most clinicians who prefer decision rules because it balances accuracy with simplicity of use and familiarity to clinicians.

Clinical Scenario—Resolution

The patient's presentation is suggestive but not clearly diagnostic of appendicitis. The clinician resorts to Alvarado's¹ clinical decision rule and calculates that the patient has 7 of 10 possible points, a positive test result with an LR of 3.1. The patient was referred for surgery, which revealed an inflamed appendix.

References for the Update

Evidence Summary and Review 1

Description of Tests and Diagnostic Standard

Two-by-two tables were constructed for each clinical characteristic found with the chart review. The 8 most accurate characteristics were used in the clinical decision rule (Table 5-7), making it one of the simplest rules available.

Main Outcome Measures

Appendicitis was diagnosed when pathologically proven. No appendicitis was defined as a normal appendicitis discovered at operation or resolution of pain without surgery. The length of follow-up of the nonsurgical patients was not defined.

Main Results

See Table 5-8.

Conclusions

Level of Evidence

Level 4.

Strengths

Although the level of evidence of this study is low, this study is noteworthy because of its early appearance in the literature and its wide acceptance.

Weaknesses

This study's retrospective design is a weakness, but it has much face validity. This study has been validated in several later studies.^1-3

References for the Evidence

Evidence Summary and Review 2

Description of Tests and Diagnostic Standard

Separate clinical decision rules were derived for men and women. The scoring system for the clinical decision rule for the men involved 7 indicators (Table 5-9); the scoring system for women involved 5 (Table 5-10). Computers were used in this study to take the data entered from a standardized form and calculate the discriminate score (DS) and compare it with the diagnostic standard.

Main Outcome Measures

The DS was compared with a diagnostic standard: appendicitis was defined as that pathologically proven. No appendicitis was defined as a normal appendicitis discovered at operation or resolution of pain without surgery. The length of follow-up of the nonsurgical patients was not defined.

Main Results

Several computer models and cutoffs were analyzed; the cutoff for the results reported in Table 5-11 was as follows. Patients with DS values below –2.00 should not have surgery, patients with a DS above –0.48 should have surgery, and patients with DS values between –2.00 and –0.48 were considered nondefined. That is, they required follow-up before the decision to operate or not was made.

Conclusions

Level of Evidence

Levels 2 and 3.

Strengths

This study had a large sample size. A standardized form was used to record all clinical data.

Limitations

This study used convenience sampling of patients that was described by the authors as “although not entirely consecutive, the series was collected by the same surgeon with regard to data collection and comprised a representative and unselected sample.”

The clinical decision rule was not validated in the original report. It has since been validated on an even larger sample of patients¹; the results were less impressive but still indicated significant power of this scoring system.

Reference for the Evidence

Evidence Summary and Review 3

Description of Tests and Diagnostic Standard

The scoring system validated in this study has been used routinely in the 2 hospitals involved in this study. A pocket chart with 10 variables and their associated scores was carried by clinicians (Table 5-12); scores suggest “consider operation,” “observe with repeated examinations,” or “observation or discharge of the patient.”

Main Outcome Measures

The diagnostic standard was positive for histologically proven appendicitis and negative for histologically disproven appendicitis or the resolution of symptoms without operation. A positive result was defined as a score of –2 or more.

Main Results

See Tables 5-13, 5-14, 5-15, and 5-16.

The overall accuracy from the receiver operating characteristic curve for the multivariate model was 0.89 for the center with consecutive enrollment vs 0.83 for the center that did not capture all eligible patients.

Using the model with a cut point of –2 or greater, as presented by the authors, produces a likelihood ratio (LR) of 5.6 (95% confidence interval [CI], 4.6-6.8) for a score of –2 or greater; when the score is less than –2, the LR is 0.31 (95% CI, 0.26-0.37).

Conclusions

Level of Evidence

Level 3.

Strengths

This study had a large sample size, and it used a standardized form on which all clinical data were recorded. The study reports a clinical decision rule that was derived and validated with good technique.

Limitations

This study reported the results from 2 centers. Most patients enrolled in the study were reported as being “consecutive.” It is difficult to assess the effect of nonconsecutive enrollment at the second hospital, which accounted for approximately 15% of the patients. However, because the overall accuracy of the score at the hospital with nonconsecutive patients was slightly worse (83% vs 89%), it is likely that the findings underestimate the true accuracy.

None of the individual clinical findings had values distinctly different from 1, allowing the clinician the opportunity to reliably rule in appendicitis. A variable with good measurement properties that decreased the likelihood of appendicitis was a normal white blood cell count (<8900/μL), with an LR of 0.16.

The investigators’ goal was to compare the predicted probability of a score by using the clinical variables with the actual outcomes. The authors recommend a cut point of –2 or greater as suggesting the need for surgery and a value –17 or less as appropriate for discharging a patient home without observation and a repeated examination. The data are presented in a fashion that allows clinicians to calculate the LR for the 3 levels of appendicitis scores. The serial LRs perform better than the dichotomous LR and match the clinical recommendations for the different levels of LRs.

Evidence Summary and Review 4

Description of Tests and Diagnostic Standard

The 10 tests evaluated were the Lindberg et al,¹ Eskelinen et al,² Alvarado,³ Fenyo,⁴ Izbicki et al,⁵ Christian and Christian,⁶ van Way et al,⁷ Teicher et al,⁸ Arnbjornsson,⁹ and De Dombal¹⁰ scores. These scores were grouped according to the population in which the score was intended for use (Table 5-17). Group A (Lindberg and Eskelinen scores) contained scores intended for use with a population with acute abdominal pain. Group B (Alvarado, Fenyo, Izbicki, and Christian scores) scores were intended for use on patients suspected of having appendicitis. Group C (van Way, Teicher, and Arnbjornsson scores) scores were derived from patients who had appendicitis. Group D (De Dombal) were scores intended for use with any patient with abdominal pain, but the diagnosis of interest was “nonspecific abdominal pain”; that is, instead of diagnosing appendicitis, it “diagnoses” pain in which surgical intervention is unnecessary.

Appendicitis was diagnosed when confirmed by pathology specimens. “No appendicitis” was defined as a normal appendix discovered at operation or resolution of pain without surgery. Patients not receiving operation were followed by telephone interview for a length of time that was undefined.

Main Outcome Measures

The collected data were used to calculate the 10 predictive scores. Patients were retrospectively assigned to outcomes that would have resulted from management that followed the score's suggestion. A 15% negative appendectomy rate is accepted as standard of care, so a score that resulted in assignments of patients leading to more than 15% was deemed unacceptable. This was done to define a minimally acceptable performance of a score.

There were 4 such criteria used for comparing outcomes from the 10 scores: (1) “[i]nitial negative appendicectomy [sic] rate” (defined as proportion of patients who did not have acute appendicitis who were assigned to the operation group), (2) “[p]otential perforation rate” (defined as proportion of patients with acute appendicitis not assigned to the operation group), (3) “[i]nitial missed perforation rate” (defined as proportion of patients with perforated appendicitis not assigned to the operation group), and (4) “[m]issed appendicitis rate” (defined as the proportion of patients with acute appendicitis who were assigned to the exclusion group).

Main Results

The prevalence of appendicitis in this study was 17%.

None of the tested scores fulfilled the criterion for an acceptable score since all had high missed appendicitis rates (Table 5-17). By calculation of sensitivity and specificity from the data provided in the study, it appears that there was a deflation of the sensitivity and inflation of the specificity in this study compared with the other attempts at validating these data, which suggests a referral bias in this or the analyzed studies. The initial diagnostic accuracy of the clinicians also did not perform at a minimally acceptable level.

Despite the disappointing performance of the scores, the investigators reported the performance of each score compared with one another. After applying each score to the entire database of patients presenting with abdominal pain (not just the populations for which the scores were intended or derived), the investigators recommended further testing of 2 scores in patients with abdominal pain or suspected of having appendicitis: the Alvarado and Eskelinen scores. The investigators also report the variables used most frequently: site and duration of pain, site of tenderness, rebound tenderness, and white blood cell count.

Conclusions

Level of Evidence

Level 1.

Strengths

This was an unprecedented, prospective, multicenter study that compared 10 clinical prediction rules for appendicitis on a single, large population at several German hospitals. The methods were fairly well described, and the criteria against which all rules were compared seemed thoughtful.

Weaknesses

The clinical database used in this study contained most, but not all, of the clinical criteria used in each clinical prediction rule. This was a complex study, and its description and tables were somewhat confusing. The division of studies into groups A and B was based on subjective data and seemed arbitrary. The included studies typically did not explain the difference between “acute abdominal pain” and “suspected appendicitis.” The data reported most thoroughly were those analyzed by groups; perhaps the most useful data presented were in the text, where groups A and B were compared on the same population.

The performance of each of the rules was surprising. The investigators provide several suggestions to explain the poor performance, mainly positive bias of the original studies and geographic differences in patient characteristics. Beyond what was explored in the discussion, the difference between the initial and actual treatment plan may explain the poor performance of the scores. Given time, the patient may lose symptoms or signs and therefore exhibit a lower score than initially recorded.

Nonetheless, it appears that the Alvarado and Eskelinen scores are the best clinical decision rules for appendicitis in patients with abdominal pain. This judgment is based on the practicality of the score and the use of the most powerful individual findings. In addition, the Alvarado rule is the oldest rule most familiar to clinicians and is the simplest to implement.

References for the Evidence

Clinical Scenario

A 24-year-old woman presents with abdominal pain, nausea, and vomiting. She describes the pain as beginning in her midabdomen 3 days ago, and it has gotten progressively worse. Her last menstrual period was 3 weeks ago and was normal; she is not sexually active. The pain has stayed in the midabdomen and not moved to other locations. On examination, she has a fever and right lower quadrant (RLQ) and rebound tenderness; her pelvic and rectal examination results are unremarkable. Laboratory evaluation reveals a left shift without leukocytosis and ketonuria.

Clinical Scenario—Resolution

The patient's presentation is suggestive but not clearly diagnostic of appendicitis. The clinician resorts to Alvarado's¹ clinical decision rule and calculates that the patient has 7 of 10 possible points, a positive test result with an LR of 3.1. The patient was referred for surgery, which revealed an inflamed appendix.