German Survey on EUS-Guided Diagnosis and Management of Gastrointestinal Stromal Tumors (GISTs) – Evidence or “Gut-Feeling”?

Abstract

Purpose: To examine practice patterns of endosonographers in diagnosing and managing gastrointestinal stromal tumors (GISTs) in Germany.

Materials and Methods: A modified published survey (Ha et al., Gastrointest Endosc 2009) was sent to endosonographic ultrasound (EUS) customers in Germany. The survey was also publicized on the homepage of an EUS interest group. To avoid duplicate opinions, participants were asked to return one survey per institution.

Results: 142 centers of roughly 850 German EUS centers responded. 25 % were from University hospitals and 74 % from community hospitals. 61 % performed > 2 EUS scans for suspected subepithelial lesions/week. Although 97 % of respondents believed that tissue acquisition with CD117 immunohistochemistry best predicts a GIST, 11 % do not perform EUS-FNA when suspecting a GIST, 68 % perform it occasionally and 18 % perform it regularly. The main EUS criteria used for a suspected GIST are the typical layer (85 %), hypoechoic appearance (80 %) and gastric location (51 %). 69 % would diagnose a GIST with negative CD117 if the EUS criteria and spindle cells are present. FNA was rated helpful in < 50 % by 55 % of participants. Size was the primary criterion for suspecting malignancy. 95 % of respondents would perform surveillance ≥ 1x/year of GISTs that are not resected.

Conclusion: There is significant variability in the diagnosis and management of GISTs in Germany. Diagnostic certainty of EUS-FNA is suboptimal in many centers and EUS is frequently used for guidance. The diagnosis of a GIST is often guided by a “gut feeling” rather than evidence. Efforts should be made to unify existing guidelines.

Zusammenfassung

Ziel: Untersuchung des praktischen Vorgehens endosonographierender Ärzte in Deutschland (GER) bei der Diagnostik und Management von gastrointestinalen Stromatumoren (GIST).

Material und Methode: Ein leicht modifizierter, publizierter Fragebogen (Ha et al, Gastrointest Endosc 2009) wurde an die Kunden von Endosonografie (EUS)-Systemen in GER versandt. Zusätzlich wurde die Umfrage auf der Homepage einer deutschen EUS-Interessensgruppe veröffentlicht. Um Doppelantworten zu vermeiden, baten wir um Rücksendung nur eines Fragebogens pro Institution.

Ergebnisse: 142 von etwa 850 EUS-Kunden antworteten, darunter 25 % aus Universitätsklniken und 74 % aus regionalen Häusern. 61 % führten mehr als 2 EUS/Woche bei Patienten mit subepithelialen Läsionen durch. Wenngleich 97 % der Teilnehmer angaben, dass eine Gewebeentnahme mit CD117-Immunhistochemie einen GIST am besten voraussagt, führen 11 % bei Verdacht auf GIST keine EUS-FNA durch, 68 % tun das gelegentlich und nur 18 % regelmäßig. Die wesentlichen EUS-Kriterien für die Verdachtsdiagnose eines GIST sind die typische Schichtenzuordnung (85 %), Echoarmut (80 %) und Magenlokalisation (51 %). 69 % würden die Diagnose eines GIST auch bei negativer CD117-Immunohistochemie stellen, wenn typische EUS-Kriterien und Spindelzellen nachweisbar wären. 55 % der Teilnehmer gaben an, die EUS-FNA sei in < 50 % der Fälle hilfreich. Tumorgröße war das primäre Kriterium für einen Malignitätsverdacht. 95 % der Umfrageteilnehmer würden GIST, die nicht reseziert werden, mindestens 1 x jährlich überwachen.

Schlussfolgerungen: Kriterien und Methoden zur Diagnose von GIST unterscheiden sich zwischen deutschen Zentren signifikant. Die diagnostische Treffsicherheit der EUS-FNA ist in vielen Zentren suboptimal und das EUS Bild entscheidet häufig über das Vorgehen. Das Management von GIST wird oft eher durch „Bauchgefühl“ als durch Evidenz gesteuert. Anstrengungen zur Vereinheitlichung von Leitlinien sind erforderlich.

Key words

abdomen - stomach - endoscopy - tumor - endoscopic ultrasound - fine-needle aspiration - subepithelial tumors - gastrointestinal stromal tumor - survey

Introduction

Subepithelial lesions (SELs) are frequently encountered as incidental findings during endoscopic procedures [1]. These lesions are often of benign nature, but they can represent malignant or premalignant conditions, such as gastrointestinal stromal tumors (GISTs). Many small GISTs likely never cause symptoms, as they can be frequently identified as incidental findings on autopsy series [2]. However, all GISTs have malignant potential and must be differentiated from other, completely benign SELs [3]. Endosonography (endoscopic ultrasound, EUS) is used for further characterization, but accuracy is poor, especially when lesions are small [4] [5]. Tissue acquisition and immunohistochemical staining are therefore considered the standard diagnostic approach to differentiate GISTs from other SELs [6]. Although there are several techniques for tissue acquisition, including fine-needle aspiration (FNA), Trucut biopsy (TCB), jumbo forceps biopsy [7] and deroofing techniques [8], as well as endoscopic [9] [10] [11] or surgical resection, an optimal standard is not available and the choice of a particular method is often based on local expertise and physician or patient preferences [12]. EUS helps to guide which technique to choose, as it can provide valuable information on the possible diagnosis, ease of tissue acquisition (e. g. location or layer) and the risk of malignancy. Which EUS findings are considered most important to practitioners for differentiating GISTs from other SELs and whether the diagnostic yield of EUS-FNA in clinical practice reflects that of clinical trials remain unclear. A recent questionnaire-based study by Ha et al. among US experts of the American Society of Gastrointestinal Endoscopy (ASGE) special interest group suggested that there are substantial variations in the diagnostic approach and follow-up of suspected GISTs [13]. Whether such variations also exist among endosonographers in European countries, such as Germany, has not been studied so far.

Purpose

The aim of this study was to analyze the management of SELs among German EUS centers, because the “EUS landscape” between the US and Germany differs. In Germany, EUS is widely used not only in academic centers but also in community hospitals, and rapid onsite cytopathological evaluation (ROSE) is not broadly available. A slightly modified German version of the US questionnaire [13] was therefore used to examine the practice patterns of endosonographers in diagnosing and managing a suspected GIST in Germany and to compare these with the previously reported US data.

Materials and Methods

Survey, Subjects and Support

A previously published survey [13] was translated into German and only slightly modified to account for recent developments in diagnostic options (Appendix 1). This survey was distributed in 2010 – 2011 via mail to all customers of leading EUS providers in Germany, namely Olympus (OLYMPUS Deutschland GmbH, Hamburg, Germany) and Pentax/Hitachi (Hitachi Medical Systems GmbH, Wiesbaden/ Germany). These companies were used as distributors in order to avoid a pre-selection by the investigators, such as targeting major academic centers or selected experts. For confidentiality reasons and to avoid possible bias, the companies were asked to only directly distribute the questionnaires to all their national customers. These companies did not provide financial support for the study. The invitation letter for the survey was also not provided by the companies, but sent under the patronage of the German Society of Ultrasound in Medicine (DEGUM) and stamped envelopes with the return address of our study center were provided by us. To increase the response rate, the questionnaire was mailed twice within one year. In addition, the survey was publicized on the homepage of the Endosonografie Club Berlin-Brandenburg (www.eus-bb.de), an EUS special interest group in the German states of Berlin and Brandenburg. In order to avoid duplicate opinions from related examiners, participants were asked to only return one survey per institution.

The study was financially supported by an educational grant from Novartis (Novartis Pharma Deutschland GmbH, Nuremberg, Germany). This company did not participate in the distribution of the questionnaire or analysis of the study. Approval was given by the local institutional review board (IRB) of the Deutsche Klinik für Diagnostik (German Diagnostic Clinic).

Statistical Analysis

For the statistical analysis of continuous variables, means and standard deviations or medians and quartiles were given. Categorical variables were described using absolute and relative frequencies. Associations between two categorical variables were assessed using Fisher’s exact test for two-by-two tables and the chi-square test. Associations with p < 0.05 are termed significant. No correction for multiple testing was performed. Therefore, only the local significance level α = 5 % is kept.

Results

Baseline Characteristics

Of an estimated nationwide 850 customers of endosonography systems in Germany, 142 centers (approximately 20 %) responded. Three-quarters of respondents were located in community hospitals and the remaining respondents came mostly from large tertiary care university hospitals ([Tab. 1]). Most referrals came from gastroenterologists, but 40 % of respondents also received their patients from primary care physicians.

Table 1 Baseline characteristics of participating German EUS centers.

answer	n [% of respondents]
practice setting of the respondents
university hospital (tertiary care)	35 [25 %]
community hospital (secondary care)	105 [74 %]
private practice	2 [1 %]
other	0
number of subepithelial lesions evaluated monthly
0 – 2	56 [39 %]
2 – 5	65 [46 %]
> 5	21 [15 %]
type of referring physicians* (multiple answers possible)
gastroenterologist	109 [77 %]
surgeon	13 [9 %]
oncologist	11 [8 %]
primary care	57 [40 %]

Total n = 142. Sum may be greater than 100 % when multiple answers were possible.

Questionnaire Responses

The criteria used for the diagnosis and management of a suspected GIST in participating German EUS centers are summarized in [Tab. 2]. Approximately half (46 %) of the respondents found the EUS appearance of a SEL in combination with FNA findings most predictive of a GIST, whereas 27 % relied on FNA-based histology and immunohistochemistry (IHC) alone. The layer of origin in the muscularis propria and a hypoechoic appearance were considered most predictive of a GIST in 85 % and 80 % of the respondents, respectively. More than half of the respondents in this German survey considered the location in the stomach as one of the most predictive criteria for the diagnosis of a GIST. The vast majority (n = 138; 97 %) of respondents considered a positive c-kit stain the most suggestive FNA feature for the diagnosis of a GIST, but 42 % of respondents were willing to diagnose a GIST based on EUS criteria if FNA results were insufficient. 69 % (n = 98) of participants would diagnose a GIST even with negative CD117 if typical EUS criteria and spindle cells are present. 55 % (n = 78) of respondents found EUS-guided FNA helpful in determining whether a SEL was a GIST in < 50 % of their cases, whereas 26 % (n = 37) of respondents found it helpful in 51 – 75 % of their cases and only 11 % (n = 16) reported FNA to be helpful in > 75 % of their cases ([Fig. 1]).

Table 2 Criteria used for the diagnosis and management of suspected GIST in participating German EUS centers.

answer	n [% of respondents]
Which features of a subepithelial lesion do you consider most predictive of a GIST diagnosis as opposed to other diagnoses? (Choose one)
EUS appearance coupled with FNA findings	65 [46 %]
FNA features (histology and immunohistochemistry)	38 [27 %]
EUS features	35 [25 %]
clinical picture	3 [2 %]
FNA features (cytology)	1 [1 %]
What EUS features of a subepithelial lesion do you consider most predictive of a GIST diagnosis as opposed to other diagnoses? (Choose all that apply)
layer of origin in the muscularis propria	121 [85 %]
hypoechoic	113 [80 %]
location in stomach	72 [51 %]
size > 5 cm	49 [35 %]
other (please specify)	13 [9 %]
What FNA features of a subepithelial lesion do you consider most suggestive of a GIST diagnosis as opposed to other diagnoses? (Choose all that apply)
c-kit stain positive	138 [97 %]
spindle cells	57 [40 %]
other (please specify)	3 [2 %]
Do you ever diagnose GIST based on EUS criteria when cytologic (or histologic/immunochemistry*) examination is insufficient for diagnosis?
yes	59 [42 %]
no	83 [58 %]

(Total n = 142). The numbers in the Table are given as percentages of respondents. The sum of the answers may not always equal 100 % because percentages are rounded.

Fig. 1 Percentage of cases in which FNA is considered helpful for the diagnosis of GIST.

Abb. 1 Prozent der Fälle bei denen eine FNA zur Diagnose eines GIST für hilfreich erachtet wird.

Despite the fact that 97 % (n = 138) of respondents believe that the cytologic or histologic demonstration of CD117 (c-kit) is the best predictor in the diagnosis of GIST, 11 % (n = 15) do not perform EUS-FNA when a GIST is suspected, 68 % (n = 96) perform it occasionally and 18 % (n = 26) perform it regularly. Of the 15 respondents who never perform EUS-FNA on a suspected GIST, 80 % found either EUS alone (n = 10) or endoscopic and clinical features (n = 2) most predictive of a GIST diagnosis. 80 % of respondents who never perform FNA on a suspected GIST were from regional hospitals (n = 12) and 20 % from university hospitals (n = 3). 67 % (n = 10) of these respondents would send all suspected GISTs to a surgeon and 27 % (n = 4) use deep biopsy techniques or endoscopic resection for a tissue diagnosis of a suspected GIST. 87 % (n = 13) of those who never biopsy a suspected GIST stated that they found FNA helpful for differentiating a GIST in 0 – 25 % (n = 9) or they did not answer this question at all (n = 4).

The EUS criteria, which are used to distinguish benign from malignant GISTs, and the criteria used to refer patients to surgery are depicted in [Tab. 3]. Three morphologic criteria were considered most important for distinguishing benign from malignant lesions, including size, extraluminal growth and the presence of lymph nodes. As an option, which was not included in the US survey, inhomogeneous echostructure was chosen by 58 % (n = 82) of respondents as a feature of malignancy. Similarly, the three dominant EUS criteria to refer a GIST for resection to the surgeon were size, extraluminal growth and the presence of lymph nodes in decreasing order ([Tab. 3]). In the German questionnaire, we added “symptoms caused by the SEL (e. g. bleeding, obstruction)”, and “at least 2 of the listed criteria” as additional options. Here, symptoms were chosen by 73 % (n = 104) of the respondents, whereas the presence of multiple criteria was important to less than 10 % (n = 13) of the respondents. Despite the possible malignant potential of GISTs, less than a quarter of respondents refer all suspected GISTs to a surgeon (20 %). The reasons why respondents would not refer a suspected or proved GIST to a surgeon were advanced age or significant comorbidities of the patient in 85 %, small size (< 2 cm), “benign” EUS criteria and lack of symptoms in 84 % and lack of histologic or cytologic criteria for malignancy in 18 % of cases. For patients with a GIST that were not referred to surgery, 50 % of respondents would perform surveillance more than once a year and 45 % would do it once a year.

Table 3 Criteria used to distinguish benign from malignant GISTs or for surgical referral in participating German EUS centers.

answer	n [% of respondents]
Which EUS criteria do you use to distinguish benign from malignant GISTs? (Choose all that apply)
size	120 [85 %]
extraluminal growth (extension beyond the bowel wall)	101 [71 %]
presence of lymph nodes	91 [64 %]
echostructure (inhomogeneous)	82 [58 %]
echogenicity (hypoechoic)	38 [27 %]
central cystic spaces	39 [27 %]
lobularity	35 [25 %]
Which features prompt a referral to a surgeon for resection of a GIST? (Choose all that apply)
size	130 [92 %]
symptoms caused by a SEL (e. g. bleeding, obstruction)	104 [73 %]
extraluminal growth	94 [66 %]
presence of lymph nodes	82 [58 %]
echostructure (inhomogeneous)	66 [46 %]
positive CD-117 (c-kit) stain	60 [42 %]
lobularity	36 [25 %]
central cystic spaces	33 [23 %]
echogenicity (hypoechoic)	25 [18 %]
at least 2 of the previously mentioned criteria	13 [9 %]
I refer all subepithelial lesions to surgeons	1 [1 %]
I refer all lesions that I think are GISTs to surgeons	29 [20 %]

Total n = 142, sum of answers may exceed 142 because multiple answers were possible.

Despite size being the most frequently mentioned criterion for malignancy, the selected cut-off values varied ([Fig. 2]). Similarly, size cut-offs for surgical referral were not uniform. Respondents chose a size of 20 mm in 42 %, 30 mm in 46 % and 50 mm in 12 % of cases as a cut-off for surgical referral. As a commentary, some respondents stated that the size cut-off was dependent on the location of the lesion within the gastrointestinal tract and on the rapidity of growth of the lesion.

Fig. 2 Specific size cutoff for suspected malignancy that was specified by 73 % (n = 88) of the respondents who chose size as a diagnostic criterion of malignancy (n = 120).

Abb. 2 Spezifische Größenangaben für einen Malignitätsverdacht, die von den 73 % (n = 88) der Teilnehmer angegeben wurden, die Größe als Malignitätskriterium auswählten (n = 120).

Influence of Practice Setting and Case Volume

In a secondary analysis, we investigated whether the practice setting, case volume and frequency of puncturing suspected GISTs affected the respondents’ answers regarding the diagnosis and management of suspected GISTs. Significant differences were only noted for the willingness to diagnose a GIST when c-kit is negative ([Tab. 4]). Almost twice as many respondents from universities as compared to community hospitals would not diagnose a GIST when cytology or histology is sufficient, but c-kit stain is negative (p = 0.0214). Despite the fact that FNA is used with a similar frequency in most institutions, it appears that resection techniques for a suspected GIST are more commonly used in university hospitals (17 %) than in regional hospitals (8 %). On the other hand, more colleagues in community hospitals (21 %) directly referred their patients with suspected GIST to surgery as compared to University centers (11 %). However, these differences were not significant. Finally, the diagnostic yield of FNA appears best for examiners who see more than 5 SELs per month and depends less on the practice setting. However, again this did not reach statistical significance ([Tab. 5]).

Table 4 Influence of practice setting on the willingness to diagnose a GIST with or without immunohistochemistry [percentage of respondents].

Question: Do you ever diagnose a GIST when cytology/histology is sufficient but the c-kit stain is negative?
practice setting	university (n = 35)	regional hospital (n = 105)	p-value
yes, always	3	12	0.0214
yes, but only if the immunohistochemistry does not suggest leiomyoma or schwannoma	49	62
no	49	26

The numbers in the table are given as percentages of respondents. The sum of the answers may not equal 100 % because percentages are rounded; p-value calculated by chisquare testing.

Table 5 Influence of practice setting and case volume on the estimated diagnostic yield of EUS-FNA [percentage of respondents].

Question: In what percentage of cases is FNA helpful in determining whether a subepithelial lesion is a GIST?
practice setting	university (n = 35)	regional hospital (n = 105)	private practice (n = 2)	p-value
0 – 25 %	29	23	50	0.7868
26 – 50 %	23	32	50
51 – 75 %	29	26	0
> 75 %	14	10	0
case volume (SELs/month)	> 5/month (n = 21)	2 – 5/month (n = 65)	0 – 2/month (n = 56)	0.6229
0 – 25 %	29	22	27
26 – 50 %	24	29	34
51 – 75 %	43	26	20
> 75 %	5	12	13

The numbers in the table are given as percentages of respondents. The sum of the answers may not always equal 100 % because percentages of unanswered questions are not included and percentages are rounded; p-value calculated by chi-square testing.

Comparison of the Results with Previously Published US Data

The distribution of respondents was significantly different (p < 0.0001) from that of the previously mentioned US study, as more respondents in the US (73 %) came from university hospitals compared to Germany (25 %). In addition, significantly more respondents in the US study evaluated > 5 SELs/month (34 %) as compared to Germany (15 %). Furthermore, 40 % of the patients in this German study were referred by general practitioners, who were a rare source of referral (2 %) in the US study (p < 0.0001). These differences must be considered, when comparing the presented data to the previous US study. The main differences of survey responses between Germany and the previous US data are subsequently discussed and also outlined in [Tab. 6].

Table 6 Significant differences in survey responses between this German study and a previously published US study (Ha et al., Gastrointest Endosc 2006)

answers	Germany(total n = 142) n [% of respondents]	US (Ha et al. GIE 2006) n [% of respondents]	p-value
What EUS features of a subepithelial lesion do you consider most predictive of a GIST diagnosis as opposed to other diagnoses? (Choose all that apply)
location in stomach	72 [51 %]	35 [26 %]	< 0.0001
Which EUS criteria do you use to distinguish benign from malignant GISTs?
echogenicity (hypoechoic)	38 [27 %]	50 [40 %]	0.0214
central cystic spaces	39 [27 %]	80 [64 %]	< 0.0001
Which features prompt a referral to a surgeon for resection of a GIST?
central cystic spaces	33 [23 %]	56 [44 %]	0.0003
What FNA features of a subepithelial lesion do you consider most suggestive of a GIST diagnosis as opposed to other diagnoses? (Choose all that apply)
c-kit stain positive	138 [97 %]	80 [60 %]	< 0.0001
spindle cells	57 [40 %]	16 [12 %]
other (please specify)	3 [2 %]	38 [28 %]
Do you ever diagnose a GIST based on EUS criteria when cytologic (or histologic/immunochemistry*) examination is insufficient for diagnosis?
yes	59 [42 %]	80 [60 %]	0.004
no	83 [58 %]	54 [40 %]

n/a = not applicable; p-values < 0.05 were considered statistically significant.

Almost twice as many respondents in this study considered the location in the stomach as one of the most predictive criteria for the diagnosis of a GIST, as compared with the US data ([Tab. 6]). US-respondents used central cystic spaces as a marker for malignancy (p < 0.0001) or as a reason for surgical referral (p = 0.0003) almost twice as often as their German colleagues. In addition, 40 % (n = 50) or respondents in the US, as compared to 27 % (n = 38) of respondents in Germany, used a hypoechoic echotexture to distinguish benign from malignant GIST (p = 0.0214). The vast majority (97 %) of respondents in Germany found a positive c-kit stain most suggestive of a GIST as opposed to other diagnoses, as compared to 60 % in the US (p < 0.0001). Sixty percent of US-respondents would consider diagnosing GIST even in the absence of a positive tissue diagnosis, as compared to 40 % of respondents in this study (p = 0.004).

Discussion

The diagnostic and therapeutic management of incidental SELs remains somewhat controversial. Although a number of diagnostic algorithms have been suggested by experts in the field, there is no consensus on the optimal management of these lesions, especially in the evaluation of suspected GISTs [1] [6] [14]. This study aimed to further define the current clinical management of suspected GISTs in Germany and to compare these findings to a previously published US survey.

The study shows that there is substantial variation of practice patterns among survey respondents. The key findings are that the vast majority (97 %) of respondents consider a tissue diagnosis with immunohistochemical staining for c-kit to be the gold standard for the diagnosis of a GIST, but nearly half of the respondents would also diagnose a GIST if the IHC is negative. Endosonographers from community hospitals were significantly more likely to diagnose a GIST without a convincing IHC as compared to their colleagues in academic centers. Although EUS-FNA was used by the majority of respondents as the primary method of tissue acquisition, 55 % reported that their yield of FNA is less than 50 % and 11 % never perform FNA if a GIST is suspected. Alternative methods such as deep biopsies or resection are used instead, the choice of which depends on the practice setting and case volume. Established EUS criteria for malignancy are used with heterogeneous frequency among respondents.

Tissue acquisition with IHC is an accepted standard to differentiate GISTs from other SELs, especially leiomyomas, and the broad acceptance of this practice is confirmed by this study and the previously published US data [13]. However, because most GISTs are located in the muscularis propria (4^th EUS layer), tissue acquisition is challenging. Jumbo biopsies or bite-on-bite biopsies have a high yield for submucosal lesions (3^rd EUS layer) but they are less effective in the muscularis propria [7]. On the other hand, endoscopic resection techniques including the muscularis propria, which might be very effective in expert hands but carry a substantial risk of perforation and can lead to tumor rupture, have been advocated lately [8] [9] [10]. Therefore, EUS-FNA is frequently chosen by the majority of endosonographers as the primary procedure of choice, possibly owing to its low risk. However, the excellent yields of 82 – 94 % for 22G-FNA [15] [16] and 76 – 86 % for 19G-needles [17] [18] in retrospective studies have not been confirmed prospectively (yield 52 – 63 %) [19] [20]. The findings in this study confirm the relatively low yield of EUS-FNA for the diagnosis of SELs in clinical practice, which was estimated to be helpful in < 50 % of cases by 55 % of the respondents. Although it appears that a few expert centers with large numbers of referrals achieve yields > 75 % for EUS-FNA of SEL, the majority of endosonographers do not share this success rate. Therefore, other methods, such as jumbo biopsies or resection techniques, are used by some. This likely explains why the combination of EUS criteria and clinical findings is used instead by a large number of respondents.

The primary EUS criteria selected for a GIST diagnosis in this study are the layer of origin and a hypoechoic appearance of the SEL. The location in the stomach was important to a majority of respondents as well. Nearly 80 % of gastric hypoechoic SELs in the 4^th EUS layer are GISTs. Approximately 60 % of GISTs occur in the stomach and only a few are found in the esophagus, which increases the pre-test probability for a GIST diagnosis if a hypoechoic SEL is located in the stomach¹. Although it can be difficult to differentiate gastric leiomyomas from GISTs based on EUS features alone, one study showed that inhomogeneous echotexture, hyperechogenic spots, a marginal halo and higher echogenicity suggest the presence of a GIST [21]. A reasonable approach might be to perform EUS follow-up for lesions appearing benign in 6 to 12 months, especially when they are small. However, further investigation is necessary if the lesion grows because this increases the likelihood of a GIST [22]. Similarly, lesions with malignant features should be considered for primary resection, especially if they are large (> 3 – 5 cm), or if EUS features suggest malignancy.

The most commonly chosen EUS criteria to suggest malignancy were size, extraluminal growth and suspicious lymph nodes in decreasing order. Although size is chosen by most respondents as a major criterion for malignancy and surgical referral, there appears to be insecurity about which cut-off to choose. More than a quarter of respondents, who considered size as a criterion for a malignant GIST, abstained from admitting to a specific cut-off. Moreover, 33 % of those who gave a specific size cut-off chose a diameter of ≥ 50 mm ([Fig. 2]). This is far outside the recommendations in available guidelines, which give cut-off values as low as 20 to 30 mm [23] [24] [25]. Therefore, substantial efforts should be made to get acceptance for the most recent guidelines. Other criteria, such as echogenicity, lobularity and central cystic spaces were infrequently chosen to suggest malignancy. The most consistent criteria for malignancy in previous studies are size and extraluminal, irregular margins [26]. Accordingly, these criteria are often used to prompt surgical referral. It is surprising, however, that lymph nodes are considered to be a marker of a malignant GIST for the majority of respondents, although GISTs rarely metastasize to lymph nodes. However, lymph nodes could suggest the presence of another malignant entity [27]. It remains unclear why central cystic spaces are used as EUS criteria for malignancy or surgical referral almost twice as often in the US as compared to Germany. Although the value of this EUS feature in the prediction of malignancy has been challenged, retrospective studies suggest that it can predict malignancy [28]. On the other hand, significantly more respondents in the US considered echogenicity as a feature to distinguish benign from malignant lesions than in Germany. Although leiomyomas tend to be more hypoechoic than GISTs, this feature is not a major criterion to distinguish benign from malignant lesions [21]. The most common reason for referral to a surgeon was the presence of symptoms. The reasons why respondents would not refer a GIST to a surgeon were advanced age or significant comorbidities, small size (< 2 cm), or “benign” EUS criteria. For patients with a GIST that were not referred to surgery, 95 % of respondents would perform close surveillance of once or more a year, which is in line with an expert opinion [1].

This study has several limitations. First, it is a based on a multiple-choice questionnaire which could limit the options for answers. However, similar to the original study by Ha et al., we left room for write-in responses [13]. In addition, direct comparison to the US study is difficult and must be interpreted against the background of different practice settings. We also added a few additional items in some questions to give a broader choice of relevant options. However, this also might affect the comparison. Another significant factor may be that many US centers have ROSE, which likely increases their yield of EUS-FNA. In contrast, only 2 % of the German centers used ROSE on a regular basis and 5 % occasionally, possibly related to cost issues [29]. Therefore, our results reflect the clinical reality in Germany rather than optimal conditions.

In summary, this study shows significant heterogeneity in the diagnosis and surveillance of incidental GISTs within German centers and points out some differences to the US. A key finding is that the diagnostic certainty of EUS-FNA is suboptimal in many centers and alternative methods, such as surgical resection, are often considered, especially with EUS criteria that place a lesion at high risk for malignancy. Close surveillance might be chosen for small lesions with the lowest yield for tissue acquisition or in patients with advanced age and poor surgical risk. The study highlights that the management of suspected GISTs is still largely directed by “gut feeling” rather than firm evidence and suggests that efforts should be made to unify recommendations of international guidelines.

Abbreviations

EUS: endoscopic ultrasound;

FNA: fine-needle aspiration;

GIST: gastrointestinal stromal tumor;

IHC: immunohistochemistry;

ROSE: rapid onsite cytopathological evaluation;

TCB: Trucut biopsy;

US: United States;

Appendix 1 Questionnaire: Diagnosis and Management of Suspected GISTs (Modified English version).

Setting of practice □ University hospital or tertiary care center □ Regional hospital or secondary care center □ Primary care or private practice □ Other (please specify)…

How many subepithelial lesions have you evaluated in the past month? □ 0 – 2 □ 3 – 5 □ > 5

From whom do you receive the majority of referrals for EUS of subepithelial lesions? □ Gastroenterologist □ Surgeon □ Oncologist □ Primary care physician

How often do you perform FNA on lesions that you suspect are GISTs on EUS? □ Always □ Sometimes □ Rarely □ Never

Which features of a subepithelial lesion do you consider most predictive of a GIST diagnosis as opposed to other diagnoses (choose one)? □ EUS criteria □ FNA criteria (cytology) □ FNA criteria (histology and immunohistochemistry: CD117 = c-kit) □ EUS criteria in conjunction with cytology/histology □ Endoscopic and clinical picture

What EUS features of a subepithelial lesion do you consider most suggestive of a GIST as opposed to other diagnoses? (Choose all that apply) □ Layer of origin in muscularis mucosa or muscularis propria (2nd or 4th EUS layer) □ Hypoechoic □ Location in the stomach □ Size > 5 cm □ Other (please specify…)

What FNA features of a subepithelial lesion do you consider most suggestive of a GIST as opposed to other diagnoses? (Choose all that apply) □ c-kit stain (CD 117) positive □ Spindle cells □ Other (please specify…)

Do you ever diagnose a GIST when cytologic/histologic/immunhistochemical examination is insufficient for diagnosis? □ Yes □ No

Which method do you prefer for gathering cytologic/histologic material from a subepithelial lesion that is suspicious for a GIST by EUS criteria? □ Deep forceps biopsy, (e. g. Jumbo forceps or button-hole technique/bite-on-bite technique) □ Endoscopic (partial) resection □ EUS-FNA (22 G or 25 G aspiration needle) □ EUS-FNA (19 G aspiration needle) □ EUS-TCB (19 G Trucut needle) □ None of the above, because I refer all suspected GISTs to a surgeon

10. Do you ever diagnose a GIST when cytologic examination is sufficient but the c-kit stain is negative? □ Yes □ No

In what percentage of cases is FNA helpful in determining whether a subepithelial lesion is a GIST? □ 0 – 25 % □ 26 – 50 % □ 51 – 75 % □ > 75 %

Which EUS criteria do you use to distinguish “benign” (very low + low risk) from “malignant” (intermediate + high risk) GISTs? (Choose all that apply) □ Size (please specify a size cut-off that suggests that a GIST to could be malignant:__________ mm) □ Echogenicity (hypo- versus hyperechoic) □ Echostructure (homogeneous versus inhomogeneous) □ Central cystic spaces □ Lobularity □ Extraluminal growth (extension past the bowel wall) □ Presence of lymph nodes

Which features prompt a referral to a surgeon for resection? (Choose all that apply and please define a size if applicable) □ Positive c-kit-stain (CD117) □ Size: > 2 cm □ or > 3 cm □ or > 5 cm ◻ □ Echogenicity (hypoechoic) □ Echostructure (inhomogeneous) □ Central cystic spaces □ Lobularity □ Extraluminal growth (extension past the bowel wall) □ Presence of lymph nodes □ At least 2 of the above criteria □ I refer all subepithelial lesions to the surgeon □ I refer all suspected GISTs to the surgeon

If you do not refer a suspected or established GIST to a surgeon, how often do you survey GISTs with EUS? □ More often than annually □ Annually □ Less often than annually □ I don’t survey them at all □ Not applicable because I refer all lesions that I think are GISTs to a surgeon

• References

• 1 Jenssen C, Dietrich CF. Endoscopic ultrasound of gastrointestinal subepithelial lesions. Ultraschall in Med 2008; 29: 236-256 ; quiz 257 – 264
• 2 Kawanowa K, Sakuma Y, Sakurai S et al. High incidence of microscopic gastrointestinal stromal tumors in the stomach. Hum Pathol 2006; 37: 1527-1535
• 3 Yang J, Feng F, Li M et al. Surgical resection should be taken into consideration for the treatment of small gastric gastrointestinal stromal tumors. World J Surg Oncol 2013; 11: 273
• 4 Hwang JH, Saunders MD, Rulyak SJ et al. A prospective study comparing endoscopy and EUS in the evaluation of GI subepithelial masses. Gastrointest Endosc 2005; 62: 202-208
• 5 Karaca C, Turner BG, Cizginer S et al. Accuracy of EUS in the evaluation of small gastric subepithelial lesions. Gastrointest Endosc 2010; 71: 722-727
• 6 Eckardt AJ, Wassef W. Diagnosis of subepithelial tumors in the GI tract. Endoscopy, EUS, and histology: bronze, silver, and gold standard?. Gastrointest Endosc 2005; 62: 209-212
• 7 Buscaglia JM, Nagula S, Jayaraman V et al. Diagnostic yield and safety of jumbo biopsy forceps in patients with subepithelial lesions of the upper and lower GI tract. Gastrointest Endosc 2012; 75: 1147-1152
• 8 Lee CK, Chung IK, Lee SH et al. Endoscopic partial resection with the unroofing technique for reliable tissue diagnosis of upper GI subepithelial tumors originating from the muscularis propria on EUS (with video). Gastrointest Endosc 2010; 71: 188-194
• 9 Zhang Y, Ye LP, Zhu LH et al. Endoscopic submucosal dissection for treatment of gastric submucosal tumors originating from the muscularis propria layer. Dig Dis Sci 2013; 58: 1710-1716
• 10 Li QL, Yao LQ, Zhou PH et al. Submucosal tumors of the esophagogastric junction originating from the muscularis propria layer: a large qstudy of endoscopic submucosal dissection (with video). Gastrointest Endosc 2012; 75: 1153-1158
• 11 ESMO/European Sarcoma Network Working Group. Gastrointestinal stromal tumors: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol 2012; 23 (Suppl. 07) vii49-55
• 12 Kang YK, Kang HJ, Kim KM et al. Clinical practice guideline for accurate diagnosis and effective treatment of gastrointestinal stromal tumor in Korea. Cancer Res Treat 2012; 44: 85-96
• 13 Ha CY, Shah R, Chen J et al. Diagnosis and management of GI stromal tumors by EUS-FNA: a survey of opinions and practices of endosonographers. Gastrointest Endosc 2009; 69: 1039-1044
• 14 Landi B, Bouché O, Guimbaud R et al. Management of gastrointestinal stromal tumours of limited size: proposals from a French panel of physicians. Dig Liver Dis 2011; 43: 935-939
• 15 Vander 3rd NootMR , Eloubeidi MA, Chen VK et al. Diagnosis of gastrointestinal tract lesions by endoscopic ultrasound-guided fine-needle aspiration biopsy. Cancer 2004; 102: 157-163
• 16 Mekky MA, Yamao K, Sawaki A et al. Diagnostic utility of EUS-guided FNA in patients with gastric submucosal tumors. Gastrointest Endosc 2010; 71: 913-919
• 17 Iglesias-Garcia J, Poley JW, Larghi A et al. Feasibility and yield of a new EUS histology needle: results from a multicenter, pooled, cohort study. Gastrointest Endosc 2011; 73: 1189-1196
• 18 Storch I, Jorda M, Thurer R et al. Advantage of EUS Trucut biopsy combined with fine-needle aspiration without immediate on-site cytopathologic examination. Gastrointest Endosc 2006; 64: 505-511
• 19 Eckardt AJ, Adler A, Gomes EM et al. Endosonographic large-bore biopsy of gastric subepithelial tumors: a prospective multicenter study. Eur J Gastroenterol Hepatol 2012; 24: 1135-1144
• 20 Polkowski M, Gerke W, Jarosz D et al. Diagnostic yield and safety of endoscopic ultrasound-guided trucut [corrected] biopsy in patients with gastric submucosal tumors: a prospective study. Endoscopy 2009; 41: 329-334
• 21 Kim GH, Park doY, Kim S et al. Is it possible to differentiate gastric GISTs from gastric leiomyomas by EUS?. World J Gastroenterol 2009; 15: 3376-3381
• 22 Lim YJ, Son HJ, Lee JS et al. Clinical course of subepithelial lesions detected on upper gastrointestinal endoscopy. World J Gastroenterol 2010; 16: 439-444
• 23 Demetri GD, von Mehren M, Antonescu CR et al. NCCN Task Force report: update on the management of patients with gastrointestinal stromal tumors. J Natl Compr Canc Netw 2010; (Suppl. 02) S1-41 ; quiz S42–S44.
• 24 ESMO/European Sarcoma Network Working Group. Gastrointestinal stromal tumors: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol 2012; (Suppl. 07) vii49-55
• 25 Nishida T, Hirota S, Yanagisawa A et al. Clinical practice guidelines for gastrointestinal stromal tumor (GIST) in Japan: English version. Int J Clin Oncol 2008; 13: 416-430
• 26 Sepe PS, Brugge WR. A guide for the diagnosis and management of gastrointestinal stromal cell tumors. Nat Rev Gastroenterol Hepatol 2009; 6: 363-371
• 27 Tokunaga M, Ohyama S, Hiki N et al. Incidence and prognostic value of lymph node metastasis on c-Kit-positive gastrointestinal stromal tumors of the stomach. Hepatogastroenterology 2011; 58: 1224-1228
• 28 Kim MN, Kang SJ, Kim SG et al. Prediction of risk of malignancy of gastrointestinal stromal tumors by endoscopic ultrasonography. Gut Liver 2013; 7: 642-647
• 29 Savoy AD, Raimondo M, Woodward TA et al. Can endosonographers evaluate on-site cytologic adequacy? A comparison with cytotechnologists. Gastrointest Endosc 2007; 65: 953-957