Symptomatic reactive hypoglycemia is highly prevalent post-esophagectomy

Summary

Impaired long-term health-related quality of life (HRQL) is a growing concern for disease-free survivors post-esophagectomy. Post-prandial discomfort, neuroglycopenic symptoms, and dumping are common and often linked to reactive hypoglycemia (RH), a condition not yet prospectively studied in this cohort. This study used continuous glucose monitoring (CGM) and detailed diaries to investigate the incidence of symptomatic and asymptomatic RH events in consecutive, disease-free patients post-esophagectomy. Disease-free patients (6–36 months post-esophagectomy, non-diabetic) were recruited. Continuous glucose readings were collected at 15-minute intervals for 7 days using a FreeStyle Libre sensor. Participants completed detailed food and symptom diaries, and validated quality of life and symptom questionnaires. Sensor data were analyzed to detect reactive hypoglycemic events, and symptoms were correlated with glucose readings. Thirty-two patients provided complete data. Almost all participants (96.9%, n = 31) had at least one blood glucose measurement <3.9 mmol/L, with 65.6% (n = 21) recording severe hypoglycemia (<3.0 mmol/L). These events are significantly more prevalent than in healthy reference populations (P < 0.001). The mean total number of hypoglycemic events (defined as interstitial blood glucose <3.9 mmol/L for <15 minutes) was 9.26 (±5.81) over 7 days. Symptomatic RH events (occurring 60–300 minutes post-meal) were experienced by 37.5% of participants (mean 1.25 ± 0.45 events), while asymptomatic events were far more frequent (mean 5.97 ± 4.32). Symptomatic RH included both neuroglycopenic (42%) and abdominal symptoms. Screening questionnaires did not accurately identify all RH cases. Importantly, RH was not associated with impaired HRQL or altered body composition. RH is highly prevalent in disease-free survivors post-esophagectomy, far exceeding general population norms. CGM is a valuable tool for diagnosis, detecting a high rate of both symptomatic and asymptomatic events. Future work should investigate whether dietary modification can reduce these clinically significant RH events.

Graphical Abstract

Graphical Abstract.

INTRODUCTION

Esophagectomy with adjuvant or neoadjuvant chemotherapy or radiotherapy is the curative treatment option for resectable esophageal cancer.^1,2 Unfortunately, esophagectomy can result in various gastrointestinal symptoms due to reconstruction of the upper gastrointestinal tract which affects transit of food through the gastrointestinal (GI) tract and gut hormonal signaling. Gut hormones, most notably glucagon-like peptide (GLP-1) and gastric inhibitory peptide, are key regulators of hunger and satiety.³ Post-esophagectomy, an exaggerated gut hormonal response which has been positively correlated with insulin release and ultimately results in symptoms and increased risk of weight loss.^3–6

Reactive hypoglycemia (RH) or late dumping syndrome (LDS) is a frequent, under-recognized post-operative complication of upper gastrointestinal surgery.³

RH is defined as post-meal hypoglycemia with autonomic nervous symptoms.⁷ LDS is the form of dumping syndrome that occurs 1–3 hours after carbohydrate consumption and is a result of an incretin-induced hyperinsulinemic response which results in hypoglycemia. This “reactive” hypoglycemia causes neuroglycopenic symptoms such as dizziness, fatigue, and feeling faint; it may also cause autonomic symptoms such as perspiration, palpitations, tachycardia, and hypotension.³ These are understood to be due to complex mechanisms including intestinal hormonal changes and a hyperinsulinemic response to food reaching the small intestine. It differs from early dumping syndrome (EDS) that occurs within 1 hour of food consumption. Symptoms of EDS are thought to be due to the rapid transportation of food into the small intestine. Reduction in gastric volume or elimination of the barrier function of the pylorus is responsible for the rapid transit. The hyperosmotic content reaching the small bowel results in fluid shifts in the plasma, which may result in symptoms such as bloating, abdominal fullness, abdominal pain, and borborygmi.⁴ Patients experiencing early and late dumping symptoms have been reported to have decreased health-related quality of life (HRQL).^8,9

LDS or RH is a well-established phenomenon post-bariatric surgery.¹⁰ Just over one-third of patients undergoing a Roux-en-Y gastric bypass surgery or sleeve gastrectomy report symptoms consistent with RH.¹¹ Reported rates of RH in the post-esophagectomy cohort range from 17% to 78%.¹² This variation in prevalence is at least partially explained, until recently, by the lack of consensus in diagnosis.¹³ Dumping syndrome can be diagnosed using symptom scoring questionnaires and provocative tests such as the oral glucose tolerance test (OGTT).³ More recently, RH after surgery has been diagnosed using continuous glucose monitoring (CGM) in combination with food and symptom diaries (FSDs).¹⁴ CGM involves applying a glucose sensor onto a body surface, usually the upper arm. Subcutaneous interstitial blood glucose measurements are recorded and saved onto the sensor which can then be transferred to a reader wirelessly.¹⁵

RH has been explored in the context of bariatric surgery and gastrectomy for gastric cancer.^15,16 To our knowledge, it has yet to be explored in the post-esophagectomy setting for esophageal cancer.

Aims and objectives

This study aims to use continuous glucose measurement and a 7-day FSD to investigate the incidence of RH in disease-free survivors at least 6 months post-esophagectomy. Furthermore, it aims to explore HRQL in patients experiencing RH.

METHODS

Participant selection and study design

Ethical approval was granted by the research ethics committee for Tallaght University Hospital and St James Hospital (2020-10 Chairman’s Action [49]).

Consecutive patients who had an esophagectomy for esophageal cancer from Jan 2018 to March 2021 with curative intent for esophageal cancer and were alive without evidence of disease recurrence on surveillance imaging for between 6 and 36 months were eligible for recruitment. Surveillance scans were performed at 6 monthly intervals. Patients with recurrence were excluded to ensure neither confounding changes from systemic anti-cancer therapy nor altered ability to eat due to disease recurrence would affect the results of the cohort study. There were no patients who had an esophagectomy for benign disease included.

Exclusion criteria included a known diagnosis of diabetes, presence of recurrence of cancer at the time of recruitment, and patients who had a previous gastrectomy to treat junctional cancers. None of the included patients were having treatment for an anastomotic stricture or had long-term enteral feeding nutrition.

Participants were contacted using phone consultation, and information packages containing a patient information leaflet, information about the FreeStyle Libre glucose sensor and a consent form were posted to each participant. Those who chose to participate were invited to an in-person clinic. All patients underwent open surgery with reconstruction using a gastric conduit via either a cervical or thoracic anastomosis, depending on primary tumor location. All patients as standard underwent a pyloroplasty to assist gastric emptying. Clinicopathological details were abstracted from the medical records. Anthropometric measurements including height (cm), weight (kg), waist circumference (cm), and hand grip strength (kg) were taken and recorded. Body mass index (kg/m²) was calculated as per the World Health Organization guidelines.

A FreeStyle Libre® glucose sensor was applied to the upper arm and the readers were activated. This system recorded blood glucose concentrations at 15-minute intervals over the 7-day period. Participants were instructed to scan the glucose sensors at least three times a day to ensure that data were recorded on the readers. No data loss from lack of scanning was identified in the cohort. Participants were provided with a sharps disposal container and were required to return the sensor and reader back to the study center.

Participants completed a series of questionnaires including the European Organization for Research and Treatment of Cancer (EORTC) esophagogastric/esophageal cancer quality of life questionnaires, which were scored according to the EORTC protocol.¹⁷ The Fatigue, Resistance, Ambulation, Illnesses, and Loss of weight scale was used to identify frailty.^18,19 The Brief Fatigue Inventory was used to measure the impact of fatigue on activities of daily living.²⁰

Maximum hand-grip strength was measured using a digital or manual grip dynamometer. Measurements were combined with gender to provide a functional parameter for sarcopenia.²¹

Participants were given an in-depth tutorial on how to populate the FSD. The FSD included the modified Oesophageal Surgery in Cancer patients: Adaption and Recovery study (mOSCAR) questionnaire, which was completed at the end of the 7-day study period (supplementary data 1).²² This facilitated identification of EDS (within 1 hour) and LDS (>2 hours post-prandially). Data from the food diaries in this study were subsequently reported and showed that the participants consumed significantly lower total amounts of energy in comparison to the general population (1794 kcal vs. 2347 kcal, P < 0.01), as well as lower amounts of protein, fat, carbohydrate, sugar, fiber, and calcium.²³

Two participants’ FSDs were not returned to the study center and thus were excluded.

Thirty-two participants were included in the study.

Study flow

Eligible patients approached (n = 89).

Declined to participate (n = 55).

Concerns regarding COVID-19 pandemic safety (n = 38).

Concerns regarding study burden/diary completion (n = 15).

Other/unspecified (n = 2).

Attended in-person clinic (n = 34).

Excluded: FSDs not returned (n = 2).

Included in final study analysis (n = 32).

Classification of hypoglycemia

Hypoglycemic events were defined as per Table 1. Hypoglycemic prevalence and glucose variability metrics were compared against age-matched healthy control subjects.²⁴

Table 1.

Classification of hypoglycemic episodes

Name	Definition	Sub-type
Hypoglycemic episode	Interstitial blood glucose ≤3.9 mmol/L for ≥15 minutes	Mild ≥3.3 to ≤3.9 mmol/L Moderate ≥3.01 to ≤3.29 mmol/L Severe ≤3.0 mmol/L
Reactive hypoglycemic episode	Reactive HG: HG occurring between 60 and 300 minutes post-prandially	Symptomatic: HG with autonomic or neuroglycopenic symptoms reported 60–300 minutes post-prandially
	Non-reactive HG: HG occurring <60 minutes or >300 minutes after a meal
		Asymptomatic: HG without any reported symptoms
Event category	Symptomatic	Asymptomatic
RH (60–300 minutes post-meal)	Symptomatic RH	Asymptomatic RH
Non-reactive (<3.9 mmol/L, >300 minutes post-meal)	Symptomatic non-RH	Asymptomatic non-RH

Statistical analyses

Data were analyzed using Microsoft® Excel® for Microsoft 365 (Version 2202 Build 16.0.14931.20648) 32-bit and International Business Machines (IBM) Statistical Package for Social Sciences (SPSS) Statistics v27 software. Descriptive statistics were performed on 32 participants. Categorical data were presented as number of participants (n) (percentage [%]), and continuous data were displayed as mean ± standard deviation, or median (interquartile range). To investigate the impact of dietary intake on RH events, independent samples t-tests were conducted to compare each nutrient with RH events. When investigating the impact of eating patterns on RH events, chi-square tests using crosstabs were performed with each eating pattern and RH events, and the continuity correction factor was used to infer significance. The one-way analysis of variance (ANOVA) function was used to compare post-prandial glucose readings between non-hypoglycemic events, asymptomatic RH events, and symptomatic RH events. In all cases, P < 0.05 was considered statistically significant.

Glucose variability metrics²⁵ calculated included mean glucose and its standard deviation. Time in range and time below range were defined as the percentage of CGM measurements between 3.9–11 and <3.9 mmol/L, respectively. Mean amplitude of glucose excursions was calculated as the average height of excursions exceeding 1 SD; continuous overlapping net glycemic action was calculated by finding the difference between glucose values at set intervals, specifically using a 60-minute interval in this study; and mean absolute glucose represented the sum of differences between successive glucose measurements divided by the total measurement time in hours (the formulae used to calculate glucose variability are in supplementary data Table 2).

RESULTS

This study included a total of 32 participants, 28 of whom were male. The majority of patients who declined to participate did so due to concerns about the safety of attending a healthcare setting as the study was carried out during the COVID-19 pandemic (n = 38) and the remainder cited concerns about the demanding nature of keeping a FSD as well as wearing a monitor (n = 15). A total 32 of 89 eligible patients attended an in-person clinic for fitting of the monitor. Demographic details are summarized in Table 2. The median age of participants was 69 years (range: 46–84 years). A number of 23/32 patients underwent a two phase esophagectomy, the remainder underwent a transhiatal (n = 7) or three phase (n = 2) in line with the unit’s usual practice.²⁶

Table 2.

Patient demographics

Variable	Participants (n = 32)
Participant demographic characteristics
Gender	Male	Female	Total
	n = 28 (87.5%)	n = 4 (12.5%)	n = 32 (100%)
Surgical history
Type of surgery
THE	7	0	7 (21.9%)
2-stage	19	4	23 (71.9%)
3-stage	2	0	2 (6.3%)
Time since surgery (months)
<12	2	0	2 (6.3%)
12–18	8	0	8 (25%)
18–24	4	0	4 (12.5%)
24–36	8	4	8 (25%)
>36	6	0	6 (18.8%)
Oncological/treatment characteristics
Tumor morphology
Adenocarcinoma	27	3	30 (93.8%)
Squamous cell carcinoma	1	1	2 (6.3%)
ASA grade
I	1	2	3 (9.4%)
II	24	2	26 (81.3%)
III	3	0	3 (9.4%)
Main curative treatment
Neoadjuvant therapy and surgery	17	4	21 (65.6%)
Surgery only	8	0	4 (12.5%)
Peri-operative chemo and surgery	3	0	2 (6.3%)

Categorical data is presented as number of participants (percentage) (n [%]). ASA, American Society of Anesthesiology; TNM, Tumor, Node, Metastasis staging system.

The median time from surgery was 26 months (range: 7–50 months). The most common morphology at histopathological analysis was the adenocarcinoma tumor type (n = 30, 93.8%). The remainder were squamous cell tumors (n = 2, 6.3%). Over two-thirds of participants were treated with neoadjuvant or perioperative treatment (n = 24, 75%). The rest were treated with surgery only (n = 8).

Anthropometry, frailty, and fatigue

Most participants included in the study were overweight or obese at recruitment, 40.6% and 12.5%, respectively. A minority were underweight (3.1%), and the remainder had a normal body mass index (43.75%). The median change in body weight (kg) since diagnosis was −9.70 (Q1 = –15.5, Q3 = –3). Weight loss of >5 kg in past 6 months was reported by 19.4% (Table 3).

Table 3.

Anthropometrics

Anthropometric data†	n = 32
Body weight, measured (kg)	74.38 ± 12.97
Height, measured (m)	1.71 (1.67–1.76)
BMI, calculated (kg/m2)	25.4 ± 3.56
BMI category‡
Underweight	1 (3.1%)
Healthy weight	14 (43.75%)
Overweight	13 (40.6%)
Obese weight	4 (12.5%)
Weight loss >5 kg reported in past 6 months‡‡
Yes	6 (19.4%)
No	25 (80.6%)
Waist circumference, measured (cm)‡‡	94.55 ± 11.77
Maximum hand grip strength, right hand (kg)‡‡	35.5 (26.4–48.6)
Maximum hand grip strength, left hand (kg)‡‡	32.6 (24.4–43.4)
§Sarcopenia (maximum grip strength <27 kg for males; <16 kg for females)‡‡,21	7 (22.6%)
Frailty and fatigue scores	n = 32
Frailty score	0 (0–1.75)
Robust health	14 (43.75%)
Pre-frail health	16 (50%)
Frail health	2 (6.25%)
¶ Global Brief Fatigue Index (BFI) score20	3.01 ± 2.02
Number of participants reporting unusual fatigue in the past week	8 (25%)

^†Data are displayed as number of participants, percentage (n [%]), mean ± standard deviation, or median (interquartile range). Abbreviations: Body mass index (BMI); Brief Fatigue Index (BFI).

^‡BMI categories: underweight range (<18.5 kg/m²); healthy weight range (18.5–24.9 kg/m²); overweight range (25–29.9 kg/m²); obese range (≥30 kg/m²).

^§Sarcopenia threshold for grip strength taken from Cruz-Jentoft et al.²¹

^¶Scores are presented on a scale of 0–10 where higher scores represent greater fatigue.

^†† n = 23 participants

^‡‡ n = 31 participants

Most participants met the robust health status criteria (43.75%). Frailty was detected in 6.25%. Strength grip dynamometer was used as an objective measurement of sarcopenia. Sarcopenia was detected in 22.6% of participants (maximum grip strength <27 kg for males; <16 kg for females).

The median global fatigue scores for participants fell into the mild category (3.01 ± 2.02).

Glucose metrics over 7-day period

The mean number of readings over the 7-day period for all participants was 597 ± 85 (433–786). The mean glucose measurement over the 7-day period was 5.91 ± 0.51 mmol/L, while the standard deviation was 1.64 ± 0.40, and the co-efficient variant was 27.74 ± 6.29%. The percentage of all values ≤3.9 mmol/L, ≤3.3 mmol, and ≤3.0 mmol/L was 4.93 ± 3.65%, 1.43 ± 1.57%, and 0.55 ± 0.62%, respectively. These events are significantly more prevalent than in age-matched healthy control populations (P < 0.001). The percentage of time in hyperglycemia was 1.98 ± 2.01% (≥11.0 mmol/L). Glycemic variability (%CV = [SD of glucose/mean glucose] × 100)²⁷ was 27.75% (Table 4²⁵).

Table 4.

Glucose variability metrics

	All participants (n = 32)
CGM use, number of readings (mean + SD) [range]	597 + 85 [433–786]
Mean glucose, mmol/L (mean + SD)	5.91 + 0.51
SD, mmol/L (mean + SD)	1.64 + 0.40
CV, % (mean + SD)	27.74 + 6.29
Percentage of glucose sensor values, median (%, IQR)
<3.9 mmol/L	4.93 + 3.65
<3.3 mmol/L	1.43 + 1.57
<3.0 mmol/L	0.55 + 0.62
>11.0 mmol/L	1.98 + 2.01
Time in range (%, range)	91.36 (70.08–98.55)
Time below range (<3.9 mmol/L) (%, range)	5.6 (0.16–23.86)
Mean amplitude of glucose excursions (range)	3.69 (1.84–6.87)
Continuous overlapping net glycemic action (range)	1.83 (0.97–2.99)
Mean absolute glucose (range)	1.77 (0.39–2.89)

Table 4 and Figure 1 report the summary of glucose metrics for participants over the 7-day study period.

Fig. 1.

Glucose metrics for all participants over the 7-day period. Mean glucose is plotted and standard deviations are displayed.

Distribution of glycemic patterns—examining the glucose measurements in combination with the food and symptoms diaries

Almost all participants had at least one blood glucose measurement ≤3.9 mmol/L (n = 31, 96.9%) and just under two-thirds (n = 21, 65.6%) had at least one blood glucose measurement ≤3.0 mmol/L.

The mean total number of hypoglycemic events per participant was 9.26 (±5.81) over the 7-day period.

The incidence and distribution of hypoglycemic events can be seen in Table 4. A total of 96.9% of participants suffered from an RH episode, defined as a blood sugar level of ≤3.9 mmol/L for ≥15 minutes, at least 60 minutes post-prandially. A total of 93.8% of participants had at least one asymptomatic hypoglycemic event, with 37.5% of participants reporting events associated with symptoms.

The mean number of asymptomatic RH episodes per participant was 5.93 ± 4.32, while the mean number of symptomatic RH episodes per participant was 1.25 ± 0.45. In terms of non-RH episodes, defined in this study as a blood sugar of ≤3.9 mmol/L at ≥300 minutes since last eating episode, 59.4% of participants experienced at least one asymptomatic non-RH event, while 6.3% experienced a symptomatic non-RH event.

The mean duration of hypoglycemic events was 43.29 minutes (±23.48 minutes). More participants (n = 30, 93.8%) experienced a daytime hypoglycemic event between 0600 and 0000 hours, while 68.8% of participants experienced a nocturnal hypoglycemic event between 0000 and 0600 hours.

The median total time spent hypoglycemic per 24-hour period was 50.63 minutes (19.42–87.86).

Symptomatic reactive hypoglycemia

A total of 16 symptomatic RH episodes were detected in 12 participants. Almost two-third (62%) of these episodes occurred during the day, while 38% of the symptomatic RH episodes were nocturnal.

Of the participants who experienced a symptomatic RH episode, 67% had these symptoms while the blood glucose levels fell within the severe category, with readings ≤3.0 mmol/L. The rest (33%) experienced symptoms during milder hypoglycemia with blood sugars between 3.3 and 3.9 mmol/L.

Neuroglycopenic type symptoms were common and reported in 42% (5/12). Symptoms reported included dizziness, slight headache, “got very shaky, felt anxious, and thought panic attack coming on,” unwell, sweating, weak, light-headed, and tiredness. The rest suffered from abdominal symptoms including an overfull stomach, discomfort, abdominal pain, bloating, and stomach cramps.

Self-reported rating of dumping symptoms

Participants completed the mOSCAR questionnaire to identify symptoms that may be related to RH or dumping. Three quarters of participants completed the questionnaire (24/32). This questionnaire is designed to detect symptoms, including neuroglycopenic symptoms and its timing post-meals (>60 and ≤180 minutes). One-third of participants (n = 8, 33.3%) who completed the questionnaire reported symptoms consistent with LDS. Over 90% (22/24) reported symptoms associated with EDS.

Symptomatic RH events were associated with greater peak blood glucose during the meal prior to the event.

The trends of the blood glucose readings can be seen in Figure 2. Mean glucose measurements were calculated 5 hours post-prandially for the following episodes: symptomatic RH (green line), asymptomatic RH (dark blue line), and non-hypoglycemic meals (teal line). The mean blood glucose measurement 60 minutes after a meal was significantly higher when the meal was followed by a RH event compared to when there was no RH event recorded (8.8 ± 2.7 vs. 7.3 ± 2.1, P < 0.02).

Fig. 2.

Comparison of mean post-prandial glucose trends after meals followed by symptomatic RH (SxMeal), meals without hypoglycemia (NonHypoMeal), and asymptomatic RH (AsxMeal). Glucose trends were measured at 15-minute intervals for 180 minutes post-prandially. The mean blood glucose measurement 60 minutes after a meal was significantly higher when the meal was followed by an RH event compared to when there was no RH (8.8 mmol/L ± 2.7 vs. 7.3 mmol/L ± 2.1).

Interstitial glucose levels are higher during meals with subsequent symptomatic reactive hypoglycemia events

Interstitial glucose levels were assessed post-prandially for meals without hypoglycaemic (HG) events, meals with asymptomatic RH events, and meals with symptomatic RH events. The mean blood glucose measurement 60 minutes after a meal was significantly higher when the meal was followed by an RH event compared to when there was no RH (8.8 ± 2.7 vs. 7.3 ± 2.1, P < 0.01) and are displayed in Figure 2.

For symptomatic RH meals, mean blood sugar trends from time 0 post-meal reveal a spike in blood sugars at approximately 60 minutes (mean peak inter-quartile range (IQR 8.8 mmol/L, ±2.7), and a subsequent dip at approximately 210 minutes (mean dip 4.6 mmol/L, range) (blue line).

For asymptomatic RH meals, the peak blood glucose readings occur at approximately 45–60 minutes post-meal and dip at approximately 105 minutes post-meal. Thereafter, the blood sugar level is maintained (teal line).

Hypoglycemic events seem to occur after a high glucose excursion and without evidence of significant numbers of fasting or nocturnal hypoglycemic events as would be evident in other causes of hypoglycemia like insulinoma.

The finding that 38% of symptomatic reactive episodes were nocturnal suggests that these were related to a very late evening meal (occurring 60–300 minutes afterwards) or possibly a delay in gastric emptying which protracted the absorption phase. True fasting nocturnal events, which would not be incretin-triggered, were infrequent.

Health-related quality of life scores

The median global QOL score of participants experiencing symptomatic RH events was compared to participants experiencing no symptomatic RH events. The overall median global HRQL score was higher for patients experiencing at least one symptomatic RH (83.33, IQR = 66.67–89.58) versus no symptomatic RH events (75, IQR = 66.67–83.33), t-test = 3.83 (P < 0.01) (Table 5).

Table 5.

Quality of life scores of study participants

Quality of life variable† scores	Participants who experienced one or more symptomatic RH event (n = 12)	Participants who experienced no symptomatic RH events (n = 20)	All participants (n = 32)	Population reference scores‡		Comparing mean HRQL scores in all participants versus
				General population	Esophageal cancer patients	General population	Esophageal cancer patients
Global quality of life	83.33 (66.67–89.58)	75 (66.67–83.33)	75 (66.67–83.33)	75 (58.3–83.3)	50 (41.7–75)	0.217 (0.830)	3.838 (<0.001)
Physical function	96.67 (73.33–100)	93.33 (86.67–100)	93.33 (80–100)	100 (86.7–100)	86.7 (66.7–93.3)	−0.674 (0.505)	2.687 (0.011)
Role function	75 (73.33–100)	100 (83.33–100)	100 (66.67–100)	100 (66.7–100)	83.3 (50–100)	−0.581 (0.565)	−2.534 (0.017)
Emotional function	91.67 (91.67–100)	87.5 (75–100)	91.67 (77.08–100)	83.3 (66.7–100)	75 (58.3–91.7)	−3.44 (0.002)	−5.112 (<0.001)
Cognitive function	75 ± 21.9	83.33 (83.33–100)	82.29 ± 22.77	86.1 ± 20	83.3 ± 21.5	−0.946 (0.351)	−0.250 (0.804)
Social function	91.67 (58.33–100)	91.67 (66.67–100)	91.67 (66.67–100)	100 (83.3–100)	83.3 (66.7–100)	−1.176 (0.248)	1.267 (0.215)
Fatigue	46.29 ± 28.36	26.11 ± 25.56	33.68 ± 28	24.1 ± 24	36.6 ± 26.6	−1.935 (0.062)	−0.590 (0.560)
Nausea and vomiting	13.88 ± 17.16	8.33 ± 20.59	10.42 ± 19.28	3.7 ± 11.7	16.8 ± 23.3	1.970 (0.058)	−1.873 (0.071)
Pain	20.83 ± 21.47	0 (0–16.67)	8.33 (0–29.17)	0 (0–33.3)	16.7 (0–33.3)	−0.906 (0.372)	−2.104 (0.044)
Dyspnea	30.56 ± 33.21	18.33 ± 27.52	22.92 ± 29.86	11.8 ± 22.8	19.2 ± 27.3	2.106 (0.043)	0.704 (0.487)
Insomnia	0 (0–0)	30 ± 38.84	22.92 ± 38.28	21.8 ± 29.7	29.8 ± 32.8	0.165 (0.870)	−1.017 (0.317)
Appetite loss	30.56 ± 41.34	20 ± 31.34	23.96 ± 35.15	6.7 ± 18.3	33.7 ± 37.4	2.778 (0.009)	−1.568 (0.127)
Constipation	0 (0–0)	5 ± 16.31	0 (0–0)	0 (0–0)	0 (0–33.3)	−0.565 (0.576)	−7.689 (<0.001)
Diarrhea	0 (0–0)	33.33 (0–33.33)	26.04 ± 14.93	7 ± 18	7.3 ± 18.9	3.314 (0.002)	3.262 (0.003)
Financial difficulty	0 (0–0)	0 (0–0)	0 (0–0)	0 (0–0)	0 (0–33.3)	0.402 (0.691)	−0.809 (0.425)

^†Data is displayed as number of participants, percentage (n [%]), mean ± standard deviation, or median (interquartile range), t-test value (Sig.[two-tailed]). Variables are scored on a scale of 0–100, where 100 represents greater quality of life/function, or higher severity/interference of symptoms.

^‡Population reference values taken from the EORTC C30 reference manual.

The median global QOL score of all participants included was 75 (IQR = 66.67–83.33) versus the general population median of 75 (IQR = 58.3–83.3). Reference ranges for esophageal cancer patients reveal a median score of 50 (IQR = 41.7–75). The one sample t-test revealed a statistically significant difference between the global QOL scores of all participants included in this study, and the general esophageal cancer population (t-test = 3.838 [P < 0.001]).

DISCUSSION

Esophagectomy, while the cornerstone of treatment with curative intent for esophageal cancer, is associated with significant and persistent nutritional issues. Challenges affecting nutritional outcomes following surgery include dumping syndrome, early satiety, and altered gastrointestinal motility,²⁸ which collectively contribute to reduced oral intake, unintentional weight loss,²⁹ and an increased risk of malabsorption and malnutrition.³⁰ The resultant nutritional decline not only impairs physical recovery and functional capacity but also has negative implications for long-term survival and overall quality of life.³¹

Elliott et al. demonstrated that this surgical procedure leads to marked alterations in the secretion of key GI tract regulatory peptides, including an exaggerated post-prandial release of hormones such as GLP-1 and blunted responses of ghrelin.⁴ These hormones are pivotal in mediating satiety, regulating gastric emptying, and influencing glucose homeostasis; and disruption of the enteroendocrine axis contributes to the profound weight loss and metabolic disturbances frequently observed in this patient group. Increased levels of GLP-1 have been correlated with decreased oral food intake, an increased incidence of dumping syndrome, and a higher prevalence of malnutrition.

RH following upper gastrointestinal surgery, such as esophagectomy, arises from the altered anatomy and physiology of the digestive tract.¹³ The prevalence is unclear, with studies reporting an incidence of 17% based on symptoms reported in clinics, which increases to 67% where dedicated symptom screening questionnaires are employed.³² The rapid passage of food, particularly carbohydrates, into the small intestine triggers an exaggerated release of gut hormones like GLP-1. This leads to excessive insulin secretion, causing a rapid drop in blood glucose levels several hours after eating and associated neuroglycopenic symptoms. The loss of normal gastric reservoir function and disrupted feedback mechanisms between the gut and pancreas contribute to this dysregulated insulin response.

While the prevalence of RH observed in our post-esophagectomy cohort is notable, it is critical to contextualize these findings against the established bariatric literature. Unlike bariatric populations—such as those undergoing Roux-en-Y gastric bypass—where the primary intervention is metabolic, our cohort underwent esophagectomy for malignancy, necessitating complex multimodal therapy, including neoadjuvant or peri-operative chemotherapy and radiotherapy. This therapeutic burden, combined with the resultant systemic physiological stress, distinguishes these patients from the elective bariatric population. Furthermore, despite significant anatomical modification of the upper gastrointestinal tract, our patients maintain intact pancreatico-duodenal anatomy. This differs fundamentally from bariatric procedures that involve direct gastric bypass or exclusion of the duodenum, which profoundly alter incretin-axis engagement. Consequently, while the reactive hypoglycemic trends share superficial similarities with post-bariatric observations, the underlying drivers in our cohort are likely influenced by a unique intersection of surgical reconstruction and possibly of long-term oncological treatment effects, warranting this dedicated study.

Incidence

A liberal definition of RH was used in this study (≤3.9 mmol/L). Almost all participants included in this study suffered from an RH episode. Severe RH (≤3.0 mmol/L) was detected in 65.6%. This is similar to the rate of severe RH (<3.05 mmol/L) reported in the people status post-Roux-En-Y gastric bypass surgery (75%).¹⁴

We chose this more liberal definition as an appropriate threshold for an exploratory study in this post-surgical population, where the true prevalence of RH is currently under-recognized and often associated with a range of symptoms, while acknowledging that our findings demonstrate severe hypoglycemia (<3.0 mmol/L) was detected in 65.6% of patients, a figure which is of undeniable clinical significance.

The mean time spent in hypoglycemia for participants was 4.93%. This is significantly greater than the 1.1% (15 minutes/day) reported time spent <3.9 mmol/L in healthy age-matched volunteers (1.1%).²⁴ The mean time spent in severe hypoglycemia ≤3.0 mmol was 0.55% (SD = 0.62), indicating that it is a quite infrequent occurrence in the post-esophagectomy cohort but is not reported in healthy control populations.

The incidence of symptomatic RH detected using CGM and FSDs combined was compared to the detection rates using the mOSCAR symptom rating questionnaire alone. CGM and FSDs detected symptomatic RH in 9/24 participants who completed the questionnaire. The mOSCAR questionnaire alone detected LDS (symptomatic RH) in 8/24 participants. Hence, the prevalence of asymptomatic hypoglycemia is greater than self-reported symptomatic screening tools would indicate.

The clear advantage of employing CGM use is in detecting asymptomatic RH cases, which was experienced by 96.9% of participants (n = 31). It also has the advantage of objectively supporting that symptoms experienced by the patients are due to hypoglycemia.

Although symptom rating questionnaires like the mOSCAR are important in defining the patient’s experience of symptoms, their disadvantage is that the symptoms reported may not be related to a hypoglycemic episode. Furthermore, due to the retrospective nature of reporting, participants may not recall all symptoms experienced. The FSDs offer an opportunity to correlate symptoms with glucose readings. A limitation of the study design is that patients were not blinded to the glucose readings and could see what the blood glucose readings were, if they so desired, potentially introducing observer bias to the symptom reporting in the diaries. As this was an exploratory study, patients were not given specific education on expected measurement ranges nor instructions on monitoring their blood glucose levels during the study other than to ensure measurements were being collected. Furthermore, there is a technical limitation regarding the reliability of current CGM technology in the very low glucose range.

Analyzing the mean interstitial glucose levels for 5 hours post-prandially revealed differential glucose trends

For symptomatic RH meals, mean blood sugar trends from time 0 post-meal reveal a spike in blood sugars at approximately 60 minutes, and a subsequent dip at approximately 210 minutes, suggesting that at approximately 3.5 hours post-meal there is a risk that the post-esophagectomy patient may experience a symptomatic dip in blood glucose level. The trends are similar to what was found in a post-gastrectomy cohort.¹⁵

The asymptomatic RH initially follows a similar trend of peak blood glucose reading occurring at approximately 45–60 minutes post-meal. The dip observed here is more gradual, occurs at approximately 105 minutes post-meal and thereafter the blood sugar level is maintained.

For the non-hypoglycemic meals, a more gradual rise in blood glucose readings in the first hour post-prandially can be observed, with less of a peak when compared to meals followed by symptomatic and asymptomatic RH events.

The apparent increase in blood glucose readings at the end of asymptomatic and non-RH meals was related to the fact that often the participant had consumed another meal or snack at that time point. This implies that counseling to reduce high-glycemic-index foods during meals may reduce the incidence of symptomatic post-prandial hypoglycemic events, an area requiring further investigation.

There is evidence in the literature that repeated exposure to hypoglycemia can escalate oxidative stress and inflammation, which may lead to damage in vulnerable brain regions, a decline in cognitive function and impaired cardiac rhythm.^33,34 In addition, repeated hypoglycemia may result in decreased caloric intake as a nutritional consequence from a fear of eating, leading to further weight loss, and malnutrition.^35,36

Interventions for reactive hypoglycemia

Several dietary strategies exist as first-line management options for RH. Advice includes consuming frequent, smaller meals throughout the day to stabilize blood sugar levels and prevent significant drops. A balanced diet is essential, prioritizing complex carbohydrates such as whole grains, fruits, and vegetables over refined carbohydrates. Lean protein should be included with each meal and snack to slow down carbohydrate absorption, and healthy fats, found in nuts, avocados, and olive oil, can further contribute to blood sugar stability. Increasing fiber intake also aids in regulating blood sugar. It’s important to limit or avoid sugary foods and drinks, as these can cause rapid spikes and subsequent drops in blood sugar.³⁷ Awareness of the glycemic index (GI) may also be beneficial, with a preference for low GI foods that have a slower impact on blood glucose levels.³⁸ If symptoms persist, some pharmacological therapies such as acarbose and somatostatin analogs can be trialled.³⁶ Surgical re-intervention in bariatric populations or enteral feeding may be necessitated.^3,13

Health-related quality of life

HRQL was investigated. No decrease in global QOL scores was detected in our sample. This study was demanding; therefore, the design of the study may have selected for patients who may have a better baseline QOL. Furthermore, the research was carried out at a high volume center for esophageal cancer care, where previous studies have shown high quality oncological outcomes.²

LIMITATIONS

Our study excluded patients with disease recurrence, which comprises a large proportion of patients post-esophagectomy. The decision to exclude patients with disease recurrence allowed for a focused analysis of metabolic outcomes in disease-free survivors, though this limits the study’s findings regarding the broader population. Future research incorporating cohorts of patients with esophagectomy for benign disease and those with disease recurrence is warranted to expand these findings. Furthermore, although patients with diabetes were excluded, there was inconsistent availability of a baseline HbA1c level or OGTT to prove the current cohort did not have diabetes.

Participants were required to populate the FSDs, the accuracy of symptoms reported and timing of the recordings could not be controlled for and may account for under-reported intake in one-third of participants.³⁸ It was challenging to define a meal in some cases, consistent with the dietary advice of “eating little and often” after esophagectomy. The overall eating patterns and the timing of food consumption were used to help differentiate meals from snacks. The requirement to document food intake and symptoms during the study period, as well as attend in-person clinic during the COVID pandemic resulting in only 34 of 89 eligible participants taking part, introducing a selection bias for more motivated or potentially more symptomatic patients. Further studies including all patients prospectively would be required to validate our findings.

Very infrequently, some participants forgot to scan their glucose sensors, meaning that glucose readings were at times interrupted but for no longer than 1 hour during the study period. The mean absolute relative difference (MARD) for CGM accuracy of the sensors used in this study was 9.2%. The MARD represents the average percentage difference between the sensor’s readings and a reference lab measurement. This is an inherent limitation of the use of continuous measurement, and the participants did not have access to an alternate method to check their blood glucose during the study.

The mOSCAR questionnaire is the standard tool used to screen for DS; however, it does not account for participants experiencing symptoms between 1 and 2 hours post-prandially. Eight participants failed to complete this questionnaire.

The gold standard to assess post-prandial glucose dynamics is the hyperinsulinemic glucose clamp, whereby a standard meal of 75 g of glucose is consumed and glucose dynamics are assessed in a controlled laboratory setting. We acknowledge that a mixed meal tolerance test offers a valuable, controlled method for validation. However, the study was intentionally designed to assess glucose dynamics in a “real-world” environment using a 7-day FSD in conjunction with CGM. Provocative tests represent an artificial scenario and a positive test with RH is often observed in patients after foregut surgery such as gastrectomy,³⁹ esophagectomy,⁴ and bariatric surgery¹⁰ without further evidence of hypoglycemic events outside of the laboratory setting.

CONCLUSIONS

Hypoglycemia is highly prevalent in this disease-free post-esophagectomy sample, and it occurs significantly more frequently than in the general population.

Symptomatic RH is relatively common after esophagectomy. CGM using a sensor is a useful tool to diagnose these events. It may also be used to identify modifiable dietary behaviors that could impact the incidence of symptomatic RH events and this is worthy of further investigation.

There is no obvious association between symptomatic RH events and a global decrease in HRQL. Our study population had better global QOL scores than the reference esophageal cancer population.

Specific author contributions: Mistura A. Kareem (Data curation, Formal analysis, Writing—original draft), Katie Byrne (Data curation, Formal analysis, Writing—original draft), Niamh Ni Mhaoinigh (Conceptualization), Jane Fagan (Data curation), Michelle Fanning (Conceptualization, Data curation, Methodology), Paul Healy (Data curation), Jessie A. Elliott (Writing—review & editing), John V. Reynolds (Conceptualization), and Claire L. Donohoe (Conceptualization, Data curation, Formal analysis, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing).

Financial support: This work was funded in part by the charitable organization Cancer Research of the Oesophagus and Stomach (CROSS).