Watermelon stomach in systemic sclerosis: its incidence and management

Authors


Prof. I. Marie, Department of Internal Medicine, Rouen University Hospital, 76301 Rouen Cedex, France.
E-mail: isabelle.marie@chu-rouen.fr

Summary

Background  To date, there are no large endoscopic studies in systemic sclerosis (SSc), and both prevalence and characteristics of watermelon stomach in SSc have not been determined.

Aims  To determine the prevalence, clinical presentation, endoscopic appearance, therapy success and long-term outcome in SSc patients with watermelon stomach and make predictions about which SSc patients are at risk for watermelon stomach.

Patients and methods  From 1990 to 2008, 264 patients were seen for evaluation of SSc. Data were collected as regards patients’ characteristics, time of watermelon stomach onset, features, therapy and outcome of watermelon stomach.

Results  Fifteen SSc patients (5.7%) exhibited watermelon stomach. SSc onset preceded watermelon stomach manifestations in 13 patients (86.7%). Most patients (86.7%) presented with iron-deficiency anaemia, two other patients experienced gastrointestinal haemorrhage. Gastroscopy disclosed typical ‘watermelon stomach’ characterized by prominent, erythematous stripes, radiating in a spoke-like fashion from the antrum to the pylorus. All patients received conservative therapy; because of deterioration of watermelon stomach, eight patients (53.3%) underwent endoscopic procedures. During follow-up, five patients (33.3%) exhibited recurrences of watermelon stomach.

Conclusions  Our series indicates that watermelon stomach should be considered when unexplained iron-deficiency anaemia occurs in SSc patients. Moreover, because watermelon stomach may be the first manifestation of SSc, patients with unexplained watermelon stomach should systematically undergo physical examination and autoantibody testing to detect the underlying SSc.

Introduction

Systemic sclerosis (SSc) is a systemic inflammatory disorder affecting the skin and other organs, particularly the gastrointestinal tract, where lesions may lead to motor activity impairment. Gastric involvement occurs in as high as 10–75% of SSc patients. It is still recognized as associated with high morbidity, as gastroparesis may result in the following: (i) inability to ingest sufficient calories, leading to malnutrition and at later stages, in cachexia and electrolyte disturbances; and (ii) gastro-oesophageal reflux exacerbation.1–10 Furthermore, gastric impairment may result in gastric bleeding related to antral vascular ectasia (GAVE) or watermelon stomach in SSc patients.5, 9, 10 Watermelon stomach affects the antrum of the stomach. At endoscopy, parallel, longitudinal and erythematous folds of gastric mucosa are seen to traverse the antrum, converging at the pylorus, resembling the stripes on a watermelon; characteristic histological features of watermelon stomach include: dilatation of mucosal capillaries with focal fibrin thromboses and fibromuscular hyperplasia of the lamina propria.5, 9–14

Currently, however, there are no large endoscopic studies in SSc and the prevalence and the characteristics of watermelon stomach in SSc have not been determined. Indeed, the aims of this retrospective study were to: (i) determine the prevalence, clinical presentation, endoscopic appearance and therapy success of watermelon stomach in SSc patients; (ii) assess both the short-term and long-term outcome of watermelon stomach in SSc patients; and (iii) make predictions about which SSc patients are at risk for watermelon stomach.

Patients and methods

Patients

From January 1990 to December 2007, 264 consecutive patients were seen for evaluation of SSc. The diagnosis of SSc was based on the American College of Rheumatology criteria.15 There were 38 men and 226 women with a median age of 59 years (range: 26–90 years). Patients were classified according to the criteria of Leroy et al.:16 90 patients (34.1%) had diffuse cutaneous SSc (dcSSc) and 174 (65.9%) had limited cutaneous SSc (lcSSc).

Patients with SSc were retrospectively included in our study if they had GAVE.

First, the medical records of all patients were reviewed and checked for patients’ characteristics, including: (i) gender, age; (ii) SSc subtype; (iii) previous medical history: cirrhosis, chronic renal failure, cardiac insufficiency; and (iv) use of treatments that might have induced/increased gastrointestinal bleeding, and notably: nonsteroidal anti-inflammatory drugs (NSAIDs), oral platelet aggregation inhibitors and anticoagulant agents.

Second, the medical records of patients were reviewed for clinical and biochemical features to assess SSc parameters associated with watermelon stomach onset, i.e. (1) systemic complications of SSc: (i) pitting scars; (ii) gastrointestinal involvement; (iii) interstitial lung disease (ILD); (iv) pulmonary arterial hypertension, characterized by pulmonary arterial systolic pressure >40 mmHg at rest on echocardiography; and (v) renal crisis; and (2) antibody status: anticentromere and anti-Scl 70 antibodies.

Watermelon stomach

The diagnosis of GAVE was determined by skilled gastroenterologists. It was based on two endoscopic appearances, i.e. (i) typical ‘watermelon stomach’ with prominent, flat or raised erythematous stripes, radiating in a spoke-like fashion from the antrum to the pylorus; and (ii) ‘honeycomb stomach’ where gastric vascular ectasia appear like a coalescence of many round angiodysplastic lesions in the antrum.

Furthermore, the medical records of patients with GAVE were reviewed for: (i) time of GAVE onset; (ii) clinical manifestations revealing GAVE; and (iii) biochemical findings at GAVE diagnosis: haemoglobin levels (g/dL), prothrombin rate and platelet count (/mm3);

Data regarding SSc patients’ GAVE therapy were reviewed as follows:

  • (i)Conservative measures, i.e. proton pump inhibitors (PPIs), iron therapy, transfusion of packed red blood cells, corticosteroids as well as cytotoxic drugs and ethinyloestradiol/norethinosterone;
  • (ii)Endoscopic therapy. Endoscopic procedures were reviewed for: neodymium: yttrium-aluminium-garnet (Nd-YAG) laser photocoagulation, argon plasma coagulation, bipolar electrocoagulation and sclerotherapy. The number of endoscopic therapy sessions was also checked;
  • (iii)Surgery, e.g. gastrectomy and antrectomy.

Finally, patients with GAVE were followed up by clinical examination and blood tests, including the haemoglobin level for at least 3 and 6 months after therapy initiation. The outcome of patients was defined as resolution and deterioration of GAVE; criteria for successful therapy were the definite disappearance of patients’ clinical signs and cessation of transfusion of packed red blood cells. Relapses of GAVE were diagnosed on the basis of clinical and/or biochemical (iron-deficiency anaemia) recurrences. The medical records of SSc patients with GAVE were further checked for causes of death.

Results

Prevalence of GAVE

From 1990 to 2007, 264 consecutive patients were seen for SSc. Of these patients, 15 who fulfilled the criteria of GAVE were identified. The prevalence of GAVE was therefore 5.7% in our population of SSc patients.

General characteristics of SSc patients

Systemic sclerosis patients consisted of 12 women and three men, with a median age of 60 years (range: 43–69 years) at the time of GAVE diagnosis.

No patient had comorbidities associated with GAVE, i.e. liver cirrhosis (alcoholic, hepatitis C or B virus-related cirrhosis), renal or cardiac failure. Furthermore, no patient received NSAIDs, oral platelet aggregation inhibitors and anticoagulant drugs. Table 1 shows the characteristics of SSc patients with GAVE.

Table 1.   General characteristics of SSc patients with watermelon stomach in our series and in the Medline literature: 1966–2008
 Our series (= 15)Medline search (= 77)18–40
  1. SSc, systemic sclerosis; lcSSc, limited cutaneous systemic sclerosis; dcSSc, diffuse cutaneous systemic sclerosis.

Gender (% of women)80%89.3%
SSc subsets60% lcSSc/40% dcSSc56.8% lcSSc/43.2% dcSSc
 dcSSc with anti-Scl 70 antibody0%9.4%
 lcSSc with anticentromere antibody66%50%
Cutaneous telangiectases60%57.1%
Telangiectases of the digestive tract
 Oesophagus15%3.9%
 Colon6.7%5.2%
 Ileon/duodenum1.3%

Time of GAVE diagnosis

Systemic sclerosis onset preceded GAVE manifestations in 13 patients (86.7% of cases). The median time interval between SSc diagnosis to the onset of GAVE was 3 years (range: 0.5–14 years). In the remaining two patients (13.3%), SSc onset was concurrently identified in association with GAVE (Table 2).

Table 2.   Time of watermelon stomach diagnosis in SSc patients and features related to watermelon stomach; findings of endoscopy at watermelon stomach diagnosis, and long-term outcome of watermelon stomach: in our series and in the Medline literature: 1966–2008
  Our series (= 15)Medline search (= 77)18–40
  1. SSc, systemic sclerosis; typical ‘watermelon stomach’ pattern with prominent, flat or raised erythematous stripes, radiating in a spoke-like fashion from the antrum to the pylorus; ‘honeycomb stomach’ where gastric vascular ectasia appear like a coalescence of many round angiodysplastic lesions in the antrum.

Time of watermelon stomach diagnosis
 After SSc diagnosis86.7%80.8%
  Median time interval between SSc diagnosis and watermelon stomach onset (range)3 years (0.5–14 years)18 months (4–108)
 Concurrent with SSc onset13.3%8.2%
 Before SSc onset10.9%
  Range interval between watermelon diagnosis and SSc onset  (4–60 months)
Manifestations revealing watermelon stomach
 Iron-deficiency anaemia86.7%90.9%
 Haematemesis/melaena14.3%10.1%
 Median level of haemoglobin (g/dL) at diagnosis (range)8.2 g/dL (4.7–10.9)6.7 g/dL (4.1–11.6)
Findings at endoscopy:
 Typical ‘watermelon stomach’ pattern100%86%
 ‘Honeycomb stomach’ pattern14%
Outcome of watermelon stomach
 Median length of follow-up (year)1.31
 Recurrence33.3%28.6%
 Death due to watermelon stomach0%1.3%

Manifestations revealing GAVE

At GAVE diagnosis, all patients exhibited anaemia; the median level of haemoglobin at GAVE diagnosis was 8.2 g/dL (range: 4.7–10.9). No patient had abnormalities regarding: platelet count and prothombin time.

As shown in Table 2, anaemia was mainly attributed to both occult and chronic gastrointestinal bleeding related to GAVE (86.7% of cases). The two other patients were admitted for acute gastrointestinal bleeding as the first manifestation of GAVE: melaena (= 1), haematemesis (= 2). Other sources of bleeding had been eliminated in all patients, using colonoscopy and gynaecological examination (for women).

Endoscopic appearance of GAVE

As seen in Table 2, in all 15 patients, gastroscopy disclosed a single endoscopic appearance, i.e. typical ‘watermelon stomach’ (Figure 1); no patient exhibited the endoscopic ‘honeycomb pattern’. Histological examination of antral biopsies revealed capillary ectasia with focal intravascular thrombi and fibromuscular hyperplasia in the lamina propria (Figure 2).

Figure 1.

 Endoscopy: typical ‘watermelon stomach’ with prominent, flat or raised erythematous stripes, radiating in a spoke-like fashion from the antrum to the pylorus.

Figure 2.

 Histological examination of antral biopsies: capillary ectasia with focal intravascular thrombi and fibromuscullar hyperplasia in the lamina propria.

Therapy of watermelon stomach

Conservative measures.  At GAVE diagnosis, 12 SSc patients previously received PPI therapy (omeprazole at a dose of 20 mg/day); in these patients, PPI regimen was increased to a dose of 40 mg/day. In the three remaining SSc patients, PPI therapy was initiated (omeprazole: 40 mg/day).

Ten patients were also given oral iron therapy. Furthermore, seven patients (46.7%) required transfusion of packed red blood cells because of severity of GAVE-related anaemia; the median number of units transfused in these patients was 6.5 units (range: 2–14).

Endoscopic therapy.  Eight patients (53.3%) still exhibited deterioration of GAVE-associated iron deficiency anaemia under conservative therapy. These patients underwent endoscopic procedures. All eight patients who underwent endoscopic procedures, concomitantly received PPI therapy, as PPI therapy favours gastric mucosal healing after endoscopic procedures.

Therapy for watermelon stomach was initiated with various endoscopic procedures: Nd-YAG laser photocoagulation (= 2), argon plasma coagulation (= 6). The median number of endoscopic sessions given to reach a stable haemoglobin level was two (range: 1–3 treatment sessions per patient). In the case of recurrence of anaemia requiring transfusion of packed red blood cells, another endoscopic session was performed until the cessation of initial symptoms. Finally, all eight patients were successfully treated with endoscopic procedures. None of them exhibited complications during endoscopic procedure of after treatment, e.g. perforations or other serious complications.

Surgical procedures.  In our patients with watermelon stomach, surgical intervention was not necessary.

Follow-up of SSc patients with watermelon stomach

The median length of follow-up in the 15 SSc patients after GAVE diagnosis was 1.3 years (range: 3 months to 9 years) (Table 2). Five patients (33.3%) exhibited recurrences of GAVE-related features: anaemia caused by chronic (= 4) and acute gastrointestinal bleeding (haematemesis) (= 1). The median time of GAVE clinical/biological recurrence onset was 10 months (range: 3–24 months) after its initial diagnosis. One patient had favourable outcome of GAVE, receiving increased regimen of PPI therapy (omeprazole: 80 mg/day). The four remaining patients, who did not improve with conservative measures, successfully underwent endoscopic therapy: Nd-YAG laser photocoagulation (= 1), argon plasma coagulation (= 3).

Finally, no patient died of GAVE. However, during follow-up, three patients died of: ILD (= 1), pulmonary arterial hypertension (= 1) and septicaemia related to bacterial overgrowth (= 1).

Patients’ features of SSc associated with watermelon stomach onset

In nine patients with GAVE (60%), SSc was typical of the lcSSc subset; anticentromere antibody was positive in six of these patients (66.7%). dcSSc was present in six patients with GAVE (40%); no dcSSc patient with GAVE exhibited anti-Scl 70 antibody. At GAVE diagnosis, the median duration of SSc did not differ between patients with dcSSc and those with lcSSc (2 vs. 3 years).

Cutaneous telangiectases were present in nine SSc patients (60%) with GAVE. Other SSc-related systemic manifestations were as follows: (i) pitting scars: (= 7); (ii) ILD (= 7); (iii) pulmonary arterial hypertension (= 2); and (iv) joint involvement (= 6); previous history of renal crisis (= 3). The characteristics of SSc patients are shown in Table 1.

Other SSc-related digestive tract impairment included: (i) telangiectases of the oesophagus (= 2) and colon (= 1); and (ii) small bowel motor dysfunction (= 3), as shown by abnormal upper gastrointestinal tract manometry, and ano-rectal involvement [faecal incontinence (= 2) and rectal prolapse (= 1)].

Discussion

First described by Jabbari et al.13 in 1984, watermelon stomach is considered to be an uncommon vascular condition. It has been observed in association with various conditions, including: atrophic gastritis, diabetes mellitus, liver cirrhosis, chronic renal failure, valvular, or hypertensive heart diseases, as well as autoimmune disorders.5, 9–14, 17 In a series of 45 patients with watermelon stomach, Gostout et al.12 have, in fact, noted that 62.2% of patients had autoimmune diseases. Such an association has subsequently been reported rarely in SSc patients.18–41 Our literature search using the Medline database (1966–2008) indeed identified 77 cases of watermelon stomach in SSc patients; these 77 cases have been reported in 22 articles and one abstract.18–40 Our series reports 15 additional cases of watermelon stomach in SSc patients. In this instance, we observed the low prevalence of watermelon stomach in SSc; in turn, among 264 nonselected patients who were referred for evaluation of SSc, a diagnosis of watermelon stomach was made in only 5.7% of cases. However, we suggest that the prevalence of watermelon stomach may be underestimated, as: (i) such subjects may be asymptomatic and iron-deficiency anaemia may be a late-stage symptom of watermelon stomach; and (ii) watermelon stomach may be misinterpreted as antral gastritis at endoscopy. Finally, only systematic endoscopic studies will allow determining the actual prevalence/incidence of watermelon stomach in SSc patients.

The time to onset of watermelon stomach has been found to be variable. The Medline review of 77 cases shows that watermelon stomach was more often diagnosed in patients who were already established as having SSc (80.8%); the median time interval between SSc diagnoses to watermelon stomach onset was 18 months. Both watermelon stomach and SSc were diagnosed concomitantly in other patients (8.2%); in the remaining patients, watermelon stomach preceded the SSc onset (10.9%).18–40 This latter finding indicates that because systemic manifestations may precede cutaneous impairment in SSc, when watermelon stomach is diagnosed, an evaluation for SSc (including clinical examination, antinuclear antibodies, nailfold capillaroscopy) should be therefore carried out systematically. Our data confirm previous authors’ findings as in 86.7% of our patients, watermelon stomach was diagnosed after the onset of SSc. Interestingly, we observed that watermelon stomach was an early manifestation in SSc, as it more often occurred in earlier phase of the disease (<5 years after SSc diagnosis). Moreover, progression of watermelon stomach and activity of SSc have rarely been reported previously; the manifestations of watermelon stomach and the activity of SSc did not seem to parallel each other (62.5%).18–40 Our findings underline that both watermelon stomach and SSc progressed simultaneously in only 33.3% of patients; in other patients, watermelon stomach occurred when other manifestations of SSc were quiescent.

Various clinical presentations may reveal watermelon stomach. In the literature analysis of 77 SSc patients, the initial presentation of watermelon stomach was more commonly recurrent anaemia related to occult gastrointestinal bleeding (90.9% of cases).18–40 In our experience, we found that 86.7% of patients presented with iron-deficiency anaemia. Our data underscore that iron-deficiency anaemia should prompt careful evaluation of SSc patients for the underlying watermelon stomach. In previous cases of SSc patients with GAVE, the median haemoglobin level at watermelon stomach diagnosis was 6.7 g/dL.18–40 Gastrointestinal bleeding has been described in the form of intermittent melaena/haematemesis in SSc patients with GAVE; in our review, acute overt digestive bleeding occurred in 10.1% of patients, i.e. melaena (= 7) and haematemesis (= 3).18–40 In our 15 SSc patients, two (13.3%) also experienced gastrointestinal haemorrhage; interestingly, in our 264 consecutive SSc patients, we found that watermelon stomach was responsible for 16% of gastrointestinal haemorrhage.

The endoscopic appearance of watermelon stomach is typically characterized by prominent, erythematous stripes radiating in a spoke-like fashion from the antrum to the pylorus; two other macroscopic gastric patterns have also been identified: (i) ‘honeycomb stomach’, characterized by a coalescence of many round angiodysplastic lesions in the antrum; and (ii) well-demarcated round- or mushroom-shaped lesion formed by a tuft of ecstatic blood vessels.17 In the analysis of 77 SSc patients, two endoscopic appearances of GAVE have been reported: typical ‘watermelon stomach’ pattern (86%) and ‘honeycomb stomach’ in the remaining patients. In this instance, all SSc patients exhibited the typical form of ‘watermelon stomach’. There is some recent suggestion that capsule endoscopy may be useful in detection of GAVE lesions;42–44 in 128 patients, with obscure gastrointestinal bleeding with normal gastroscopy, capsule endoscopy demonstrated GAVE in 4.7% of cases.42 Nevertheless, no definite conclusion can be drawn from these results and this study warrants further investigation.

From a practical point of view, knowledge of predictive factors of watermelon stomach onset appears essential to improve the management of SSc patients. In our review of 77 cases of SSc and in the present series, watermelon stomach more often occurred in women (>80% of cases); these findings are not unexpected, as SSc is more frequently encountered in women. Watermelon stomach has been mentioned in patients with dcSSc and lcSSc. In our review of 77 patients, watermelon stomach tended to be equally associated with lcSSc (56.8%) and dcSSc (43.2%).18–40 In our 264 SSc, the prevalence of watermelon stomach was similar in patients with lcSSc and dcSSc (5.2% vs. 5.6%). Our review of 77 SSc patients also shows that GAVE tends to occur earlier in dcSSc than in lcSSc; the median duration of dcSSc was shorter before watermelon stomach onset compared with that of lcSSc (9 vs. 12 months).18–40 This study confirms these data, as our dcSSc patients exhibited watermelon stomach shorter to SSc diagnosis compared with lcSSc patients (2 vs. 3 years). Moreover, in the literature review, we found that 9.4% of dcSSc patients exhibited anti-Scl 70.18–40 In this instance, we noted that no dcSSc patients had anti-Scl 70 antibody; concomitantly, we found that about half of 249 remaining SSc patients, without GAVE, exhibited anti-Scl 70 antibody (P < 0.05). This latter finding suggests that anti-Scl 70 antibody may be a negative predictive parameter of watermelon stomach in patients with dcSSc, although no definite conclusion can be drawn from our data.

The exact pathological mechanisms of watermelon stomach still remain unknown in SSc. Previous authors have postulated that:

  • (i)Both antral mucosal prolapse through the pylorus and disordered gastric muscular motility may result in submucosal ischemia and both elongation and dilatation of mucosal vessels.9–14 The hypothesis is strengthened by the histological features of fibromuscular hyperplasia of the lamina propria and dilatation of the mucosal capillaries. Previous manometric studies have provided evidence for the occurrence of high amplitude, uncoordinated gastric antral contractions in SSc patients, which could result in forces theorized to be responsible for GAVE.9–14 In this instance, 13.3% of our patients exhibited intestinal pseudo-obstruction;
  • (ii)Watermelon stomach may be part of SSc-associated diffuse cutaneous telangiectatic lesions.18–40 Our review of 77 SSc with GAVE shows that patients concomitantly exhibited telangiectases involving: (i) the skin (57.1%); and (ii) the digestive tract: oesophagus (3.9%), duodenum (1.3%), ileon (1.3%), colon (5.2%).18–40 Previous authors have also described two cases of rectal vascular ectasia akin to the endoscopic appearance of the antrum in SSc patients.45, 46 Our 15 SSc patients also presented with telangiectases involving the skin (60%) and the digestive tract (oesophagus: = 2, colon: = 1). Our findings tend to support the theory of a diffuse SSc-related vasculopathy responsible for each of the lesions.

Optimal therapy for management of watermelon stomach remains unclear.47, 48 Indeed, we failed to find randomized and controlled studies evaluating nonsurgical therapy in patients with GAVE; these data may be, probably, explained by the rarity of GAVE making comparisons of treatment rather impossible.

General supportive care includes iron replacement therapy, identification and treatment of iatrogenic coagulopathy and avoidance of substances that might cause mucosal damage and/or bleeding.47, 48 Furthermore, patients are routinely given standard doses of PPIs (20–40 mg/day).18–40 In addition, transfusion of packed red blood cells may be required in patients with GAVE who exhibit: acute gastrointestinal bleeding and/or poor clinically tolerated anaemia. In our review of 77 SSc patients with GAVE, the median number of transfusion of packed red blood cells was four.18–40 Our data confirm these previous results, as the median number of transfusion of packed red blood cells was 6.5 in our patients. Other alternative treatment modalities have been proposed in patients with GAVE; however, there were mainly case reports, i.e.:

  • (i)Steroids. In our review of 77 SSc, we have found that one patient was given steroids (75 mg/day) successfully; another patient received combined therapy of intravenous methylprednisolone and cyclophosphamide. In this later patient, repeated endoscopy was normal after the 3-month therapy.18–40 Nevertheless, no definite conclusion can be drawn from these two case reports’ findings, and the use of such toxic regimens may not be justified in SSc patients with GAVE, in the absence of life-threatening co-existing complications of SSc;
  • (ii)Hormonal therapy. In our analysis of 77 SSc, two patients received ethinylestradiol (50 μg) and norethisterone (1 mg), resulting in clinical improvement in one patient. Although, hormonal therapy did not modify the endoscopic appearance in this later patient 18–40; therefore, it is likely that bleeding will recur on discontinuation of hormonal therapy. In fact, the use of long-term hormonal therapy should be carefully weighted against the risks; recent meta-analyses have indeed shown the potential risks of long-term hormonal therapy in women, resulting in increased risks of cardiovascular diseases and cancer;
  • (iii)Octreotide has an inhibitory effect on angiogenesis. In three non-SSc patients with GAVE, octreotide has been found to stop bleeding (= 1) and decrease transfusion requirements (= 2).49 In our review, we failed to find cases of SSc patients with GAVE receiving octreotide. One of our patients, who exhibited intestinal pseudo-obstruction, received octreotide therapy, which was unsuccessful for watermelon stomach improvement.
  • (iv)Other pharmacological therapies (H2 histamine antagonists, calcitonin, tranexamic acid, alpha-interferon, serotonin antagonists, thalidomide) have also been used as treatments for watermelon stomach in case reports of non-SSc patients with variable success.47, 48, 50

Previous investigators have mentioned that 60–70% of non-SSc patients with GAVE are transfusion-dependent because of recurrent anemia.18–40 Endoscopic therapy is indicated when clinically significant gastrointestinal bleeding occurs or when occult bleeding with resultant anaemia becomes refractory to conservative therapy, i.e. thermal-contact procedures, sclerotherapy and thermal-noncontact modalities [neodymium-yttrium-aluminum garnet (Nd:YAG) laser, argon plasma coagulation]. Both Nd:YAG laser and argon plasma coagulation are endoscopic treatments of choice in patients with GAVE; such therapies have been found to be effective in 75–100% of patients.51–53 In our review of 77 SSc, 64.9% of patients successfully underwent endoscopic treatments: (i) Nd:YAG laser (80%), argon plasma coagulation (10%) with a median number of sessions per patient of 5.5; and (ii) bipolar electrocoagulation (4%) and injections sclerotherapy (6%).18–40 Our findings confirm previous authors’ data; endoscopy therapy was, in fact, performed in 53.3% of our patients: argon plasma coagulation (75%) and Nd:YAG laser (25%); endoscopic procedures proved successful in all these patients. Finally, a pilot study has recently suggested that endoscopic mucosal ablation, by using the HALO90 system, may be a helpful treatment option for chronic bleeding related to watermelon stomach.54

Until now, no randomized trials have compared the different surgical procedures in patients with GAVE, antrectomy being by far the most commonly used procedure.47, 48 Surgery is recommended in patients with GAVE who failed treatment with endoscopic measures.18–40 Nevertheless, surgery has significant morbidity and mortality as most of elderly patients have co-morbid illnesses; previous reports suggested a 7.4% operative mortality in non-SSc patients with GAVE.17 In our review of 77 SSc, 13% of patients had to undergo surgical procedures, i.e. antrectomy (= 5), Billroth I gastrectomy (= 3), subtotal gastrectomy with Billroth II anastomosis (= 1) and both antral and pyloric involvement (= 1).18–40 In our series, no SSc patient with watermelon stomach underwent surgery.

In the review of 77 SSc patients, the median length of follow-up was 12 months; 28.6% of patients developed recurrences of GAVE, the median number of relapse being one (range: 1–4).18–40 In our experience, a higher number of SSc patients experienced a recurrence of watermelon stomach (33.3%) within 3–24 months after its initial diagnosis. Our data underline that SSc patients require long-term follow-up of watermelon stomach to detect its relapse at an early stage. In our five SSc patients who exhibited recurrence of watermelon stomach, course was favourable after initiation of conservative measures (20%) and endoscopic procedures (80%). Finally, in our review of 77 SSc, 6.4% of patients died of GAVE-related gastrointestinal bleeding (= 1), cerebral infarction after partial gastrectomy (= 1) and SSc-related renal crisis (= 1).18–40 In our series, no patient directly died of watermelon stomach.

In conclusion, watermelon stomach should be considered when both chronic and unexplained iron-deficiency anaemia and/or gastrointestinal haemorrhage occur(s) in patients with SSc. Moreover, physicians should be aware that such gastrointestinal bleeding from watermelon stomach may occur any time during the course of SSc. In contrast, watermelon stomach may be the first manifestation of SSc; in essence, patients with unexplained watermelon stomach should undergo systematically directed questioning, complete physical examination and autoantibody testing to optimize both early diagnosis and management of underlying SSc.

Acknowledgement

Declaration of personal and funding interests: None.

Ancillary