Background and aims The natural history and growth pattern of pancreatic serous cystic neoplasms (SCNs) are not well understood. This study was designed in order to get insight into the growth rate of SCNs and to suggest recommendations for their management.
Methods Patients with well-documented incidentally discovered or minimally symptomatic SCNs who underwent yearly surveillance MRI were analysed using a linear mixed model. The growth rate and the effects of different fixed factors (sex, personal history of other non-pancreatic malignancies, radiological pattern, clinical presentation, tumour site) and random factors (age and tumour diameter at the time of diagnosis) on tumour growth were investigated.
Results Study population consisted of 145 patients. Estimated overall mean growth rate was 0.28 cm/year, but the growth curve analysis showed a different trend between the first 7 years after the baseline evaluation (growth rate of 0.1 cm/year) and the subsequent period (years 7 to 10, growth rate of 0.6 cm/year, p<0.0001). Tests for fixed effects demonstrated that an oligocystic/macrocystic pattern and a personal history of other tumours are significant predictors of a more rapid mean tumour growth (p<0.0001 and 0.022, growth rates of 0.34 cm/year). Furthermore, tumour growth significantly increased with age (p=0.0001).
Conclusion Overall, SCNs grow slowly, and an initial non-operative approach is feasible in all the asymptomatic or minimally symptomatic patients. The oligocystic/macrocystic variant, a history of other non-pancreatic malignancies and patients' age impact on tumour growth. In any case, a significant growth is unlikely to occur before 7 years from the baseline evaluation. Tumour size at the time of diagnosis should not be used for decisional purposes.
- serous cystic neoplasms
- follow-up studies
- pancreatic surgery
- pancreatic tumours
- pancreatic cancer
- abdominal MRI
- acute pancreatitis
- biliary duct carcinoma
- pancreatic pseudocyst
Statistics from Altmetric.com
- serous cystic neoplasms
- follow-up studies
- pancreatic surgery
- pancreatic tumours
- pancreatic cancer
- abdominal MRI
- acute pancreatitis
- biliary duct carcinoma
- pancreatic pseudocyst
Significance of this study
What is already known about this subject?
Serous cystic neoplasms of the pancreas should be resected when symptomatic or in case of an unclear differential diagnosis with mucin-producing tumours.
It has been proposed that asymptomatic lesions larger than 4 centimeters should be also resected (in surgically fit patients) because of a very rapid growth rate.
Small asymptomatic lesions can be observed with periodic imaging.
What are the new findings?
Serous cystic neoplasms of the pancreas grow slower than previously thought.
A significant growth is unlikely to occur before 7 years from the baseline evaluation.
Maximum diameter at the time of diagnosis is not a predictor of tumour growth.
Macrocystic pattern, a history of other non-pancreatic malignancies and age are significant predictors of tumour growth.
How might it impact on clinical practice in the foreseeable future?
Asymptomatic or minimally symptomatic serous cystic neoplasms larger than 4 cm can be safely managed by the same periodic surveillance protocol.
Patients presenting with factors that impact on tumour growth are more likely to require resection at the long term.
Semiannual or annual follow-up is not necessary for asymptomatic or minimally symptomatic neoplasms; a time frame of at least 2 years seems appropriate.
In the last 2 decades, there has been an increased awareness of serous cystic neoplasms of the pancreas (SCNs), mainly due to the widespread use of cross-sectional imaging.1 2 These lesions are virtually always benign neoplasms, but local growth can result in abdominal symptoms.3–5 The principal diagnostic effort is directed towards differentiating SCNs from the potentially malignant mucin-producing lesions (such as mucinous cystic neoplasms and intraductal papillary mucinous neoplasms (IPMNs)). Radiological characteristics that may help to identify SCNs include the presence of multiple cysts measuring 2 cm or smaller separated by fibrous septa that may coalesce into a central scar, although macrocystic variants have been described.6–8
While resection is mandatory in case of symptomatic lesions or when differentiation from a mucinous neoplasm is uncertain, there is little information to predict if an asymptomatic SCN will grow enough to cause symptoms, and this greatly influences the decision on whether to resect the lesion. It has been proposed that tumours larger than 4 cm should be resected (regardless of symptoms) due to a more rapid growth rate.9 However, the reason why large tumours appear to have a faster growth rate remains unclear, and it is similarly uncertain whether this has an impact on malignant potential. Such aspect is of great importance because an operative approach may expose asymptomatic patients to major surgery, risk of postoperative morbidity, mortality and possible long-term impairment of pancreatic function with unclear benefit.
This study analyses a large cohort of patients with well-documented SCNs who were followed yearly with serial MRI. A linear mixed-effects model with random effects was employed to investigate the growth rate and the factors that may have an effect on tumour growth and—ultimately—to understand the natural history of these neoplasms.
Patients and methods
This study was approved by the Institutional Review Board. All patients with an SCN in follow-up at the authors' institution were considered for inclusion in this study. Inclusion criteria were as follows: (1) Certain radiological diagnosis: MRI with MR cholangio-pancreatography (MRCP) was employed as the standard, high-resolution imaging modality. Typical features of SCNs included the presence of high-signal-intensity microcysts on T2-weighted images, and the additional finding of a central dark scar was considered pathognomonic.6 7 The lack of a communication with the main pancreatic duct at MRCP also helped to distinguish SCNs from the more septated branch-duct IPMNs. Because the imaging aspects on MRI related to oligocystic and macrocystic forms are non-specific, in patients in whom the differentiation from mucinous cystic neoplasm was uncertain or when it was unclear if there was communication with the main pancreatic duct, endoscopic ultrasonography (EUS) with aspiration of cyst fluid was carried out. Only lesions with low (<5 ng/ml) carcinoembryonic antigen levels and aspirates classified as consistent with SCNs were included in the present analysis. (2) Availability of serial MRI: According to our institutional protocol, MRI was then performed on a yearly basis. Patients with at least two studies were included in the analysis. Maximum tumour diameter was recorded at the diagnosis. Thereafter, the maximum tumour diameter was measured in the same scan plane of the index MRI examination. In multifocal SCNs, the largest lesion was measured. MRIs were reviewed by a single expert pancreatic radiologist (RM).
Data analysis and statistics
Demographic and clinical details were retrieved from our electronic database, while follow-up information was obtained through our pancreatic outpatient clinic. The growth rate of SCNs and the effects of different factors on tumour growth were analysed by employing a linear mixed-effects model. This model is based on restricted maximum likelihood method and handles correlated data with unequal variance (ie, repeated measurements on each subject) as well as unbalanced designs (ie, studies with an unequal number of repetitions between subjects). The dependent variable was tumour growth (vector of maximum tumour diameters on MRI performed at any given time point).
Response from a subject (tumour growth) is thought to be the sum of so-called fixed and random effects: fixed effects are factors for which the only levels under consideration are contained in the coding of those effects; random effects are factors for which the levels contained in the coding of those factors are a random sample of the total numbers of levels in the population for that factor.
Fixed effects were sex (male or female), personal history of other non-pancreatic malignancies (yes or no), tumour presentation (asymptomatic or minimally symptomatic), radiologic pattern (microcystic or oligocystic/macrocystic) and tumour site (head–uncinate process, body–tail or multifocal). Age and maximum tumour diameter at the time of diagnosis were considered to be random effects and were handled as covariates. Random effects were fitted by introducing an ‘effect by subject’ interaction term to account for possible subject variations.
Type III tests for fixed effects and covariance parameters were used for the interpretation of potential predictors of tumour growth. In addition, univariate pairwise comparison of estimated marginal means across levels of fixed effects was made. Estimated marginal means are predicted means that are calculated from the fitted model and are adjusted appropriately for any other variable. Tumour growth rate was extrapolated from estimated marginal means and was represented using line plots of mean tumour diameters across levels of fixed effects. The difference in the overall growth trend between adjacent time points was assessed using custom hypothesis testing (contrast coefficients). Growth curves were truncated at 10 years. All tests were considered statistically significant for a p value <0.05. SPSS software (rel. 19) was used for statistical analysis (SPSS Inc., an IBM company, Chicago, Illinois, USA).
One hundred and forty-five patients were included in the study, the characteristics of whom are summarised in table 1. Mean age was 60.7 years; mean tumour diameter at the time of diagnosis was 28.6 mm (SD=13.4, range 6–80). In those patients with minimal symptoms, symptoms did not correlate with the presence of a pancreatic mass and did not affect the patients' quality of life (mild sense of fullness or dyspepsia responsive to medical treatment). A non-operative policy was discussed with these patients, and all gave their consent for this approach. Sixteen patients had a medical history of other non-pancreatic malignancies. Five had breast cancer, three colonic cancer, three papillary cancer of the thyroid, two endometrial cancer, and one each with ovarian cancer, renal cancer and leukaemia (in remission after chemotherapy).
An EUS-guided aspiration of the cyst fluid was needed to establish a diagnosis in 16 patients in whom the radiological picture was unclear. Aspirate characteristics consistent with a SCN included the presence of loose clusters or monolayered sheets composed of cuboidal cells with indistinct cell borders and granular or clear cytoplasm. Nuclei were small, round, with fine chromatin and indistinct nucleoli, and devoid of mitotic activity. Mean carcinoembryonic antigen levels were 2.3 ng/ml (range 1–44).
Median interval between the first and the last study was 84 months (range 24–180); the majority of patients had at least six serial MRIs (see supplementary material). The linear mixed model estimated a mean growth of 0.28 cm/year. However, the line plot analysis suggested that tumour growth is very slow in the first 7 years from the baseline evaluation, while it substantially accelerates thereafter (figure 1). This was confirmed by the custom hypothesis test (estimated mean growth for years 1 to 7=0.1 cm/year, estimated mean growth for years 7 to 10=0.6 cm/year, p<0.0001). An oligocystic/macrocystic radiologic pattern on MRI (p=0.0001) and a personal history of other non-pancreatic malignancies (p=0.022) were found to be significant predictors of tumour growth in the tests for fixed effects. From the covariance parameters (Wald statistics), we can deduce that tumour growth increases with age (p=0.0001), while diameter at the time of diagnosis has no significant effect on tumour growth (p=0.475). Results from tests of fixed effects (with estimated tumour growth rates) and from tests of random effects are outlined in table 2. Line plots of the estimated marginal means stratified by significant fixed effects are shown in figure 2 (see supplementary material for non-significant factors).
According to the assumptions of a linear mixed model, the effects of significant predictors of tumour growth are cumulative. This aspect, however, could not be strictly verified due to the small number of combined events observed, but a more rapid tumour growth when two significant factors were present may be speculated.
One hundred and twenty-two out of 145 patients (84.1%) remained in follow-up, while 23 (15.9%) underwent a pancreatic resection during this period of surveillance. Median time to resection was 48 months from the baseline evaluation (range 12–132). Among the 23 patients who underwent resection, four (all with microcystic lesions) chose resection despite a minimal tumour growth. In the other 19 patients, the main indication for resection was an increase of tumour size. Concomitant abdominal symptoms developed in eight patients. Overall, in 14 of 19 patients, a factor with a fixed effect on tumour growth (according to the model shown here) was present. In particular, 10 patients had an oligocystic/macrocystic SCN, and four had a history of other non-pancreatic malignancies. In four patients, a constant tumour growth occurred (resection was undertaken at 1, 2, 4 and 5 years, respectively), while in the remaining patient, a rapid increase in growth rate was observed between the ninth and the tenth year from the time of diagnosis. The resection rate of oligocystic/macrocystic lesions (10/21, 47.6%) was greater than that of the microcystic counterpart (9/120, 7.5%, excluding the four patients who chose resection), with a p<0.0001 (Fisher exact test) and an OR of 11.21 (95% CI 3.75 to 33.45). Similarly, the resection rate of patients with a history of other non-pancreatic malignancies (4/16, 25%) was greater than that of other patients (15/125, 12%), but this did not reach statistical significance (p=0.232, 95% CI 0.69 to 8.56, Fisher exact test). Mean baseline tumour diameter in resected patients was only 28.74 mm (SD=13.5, range 15–80). Pathological analysis confirmed the diagnosis of SCN in all the resection specimens. There were no cases of serous cystadenocarcinoma.
As with any other pancreatic cyst, SCNs are being detected with increasing frequency due to the widespread use of advanced cross-sectional imaging modalities.1 2 A correct radiologic diagnosis requires familiarity with the morphologic spectrum of these lesions, and the presence of various atypical elements or confounding morphology with mucin-producing neoplasms of the pancreas may be a source of diagnostic uncertainty. Furthermore, the additional value of EUS with cyst fluid analysis is somewhat controversial because previous studies describing the cytological features of SCNs and other cystic lesions of the pancreas are limited and demonstrate diagnostic accuracies ranging widely from 10% to 100%.10 11 Surgical resection must be considered in patients with an inconclusive differential diagnosis with a mucinous neoplasm and for well-documented symptomatic SCNs. In contrast, no definitive treatment recommendations regarding asymptomatic lesions currently exist. There is little information available in the literature for predicting whether an SCN will grow large enough to cause symptoms, how rapidly SCNs grow and whether some morphologic or patient-related features can be used for predicting biological behaviour of individual SCNs. This dilemma represents a substantial clinical issue because pancreatic surgery is not without mortality and is still associated with high morbidity rates and because long-term sequelae of loss of pancreatic tissue and function are not negligible.12 In an attempt to define the growth potential of SCNs, Tseng et al analysed a large group of patients and reviewed serial radiographs in 24 subjects, with a median interval of 23 months between the first and last study. Tumour size at the time of diagnosis was found to be associated with symptoms, which were present in 22% of patients having lesions <4 cm and in 72% having lesions >4 cm (p<0.001). The cut-off of 4 cm was used to calculate growth curves, and a significant difference was observed between the two groups (0.12 cm/year for patients with tumours <4 cm versus 1.98 cm/year for patients with lesions >4 cm, p<0.001). Overall, the median increase in size was 0.6 cm/year. As a result, the authors recommended resection for surgically fit patients with SCNs larger than 4 cm, regardless of symptoms.9 However, more than half of the sample (14/23) had only two serial studies, six patients had three studies, and only four had more than three studies. In another work by Allen et al on the selective approach to the resection of pancreatic cystic neoplasms, tumour growth was calculated in a group of 10 patients with pathologically proven SCNs. A similar overall growth rate was found (0.5 cm/year), without any association between the size of the lesion and the rate of growth.13 Lastly, a recent study by Fukasawa et al determined tumour growth in 18 patients with serial radiographs (median follow-up 58 months) to be 0.29 cm/year, and no significant factor associated with tumour growth was found. In these two latter studies, resection was recommended for patients with symptoms, for healthy patients in whom substantial growth had occurred and when the diagnosis was uncertain.14
Given the paucity of studies in the field and the small number of patients analysed, information on tumour growth has to be interpreted with caution. The present study analysed a group of 145 patients with well-documented SCNs (mean diameter of the diagnosis of only 28.6 mm) who underwent periodic surveillance imaging studies. A point of clinical importance is that only patients who were followed up with the same imaging modality (MRI + MRCP) performed on a yearly basis were analysed. Maximum tumour diameter was objectively recorded in the same scan plane of the baseline examination, growth rate was measured, and potential predictors of the growth itself were investigated. As already pointed out by other authors, one could observe that any assessment of tumour growth is limited by the validity of the means or mathematical model used to calculate tumour size.9 Diameter-based curves are commonly accepted for the estimation of tumour size and growth, although more complex formulas (based on volume rather than on maximum diameter assessment) can be used. Arguably, tumour volume and doubling time may have less applicability in cystic lesions than in solid tumours.
The authors also acknowledge that a pathologic diagnosis was obtained only in a fraction of patients who underwent either EUS with aspiration (16 subjects, 13.0%) or resection (23 subjects, 15.9%). However, all the resected specimens were SCNs (and 10 neoplasms out of 23 were macrocystic), and the criteria used for diagnosis of SCNs in the other patients without histologic confirmation are well accepted. Study of tumour growth was carried out by employing a linear mixed model with random effects. This design has the advantage of allowing observation of change over time and of controlling individual differences in parameters of relevance. Furthermore, the mixed model does not require the data set to be balanced, and so all patients with at least two measurements could be included in the analysis. Of note, the majority of patients had at least six serial MRIs, and the median interval between the first and the last study was 84 months (range 24–180). This is a substantially greater follow-up than in previously published studies. Age and tumour diameter at the time of diagnosis were considered as random effects and handled as covariates, providing a greater ability to investigate the association between these two continuous factors and tumour growth. Pairwise comparison of the estimated marginal means was carried out among specific fixed factors (sex, history of other non-pancreatic malignancies, tumour presentation, radiologic pattern and tumour site) and then plotted for interpretation of effects. Marginal means are predicted means, not observed: in the linear mixed model, the dependent variables are equal to linear combinations of fixed and random effects plus an error term. By plotting the estimated means, an analysis of the effects without the error is obtained (while plotting the observed means across levels of a factor would be the same as plotting the predicted values plus the errors).
The main finding of the present study was that the estimated mean growth rate of SCNs is 0.28 cm/year, substantially different from the reports by Tseng et al and Allen et al9 13 but similar to the observation made by Fukasawa et al.14 According to the line plot shown in figure 1, mean tumour growth seemed to be very slow for the first 7 years after the baseline evaluation and more rapid from the seventh year and onwards. Such difference was confirmed to be statistically significant by the custom hypothesis test: the estimated mean growth in the first 7 years was in fact only 0.10 cm/year, and then the growth rate raised at 0.60 cm/year (p<0.0001).
In the tests for fixed effects, the oligocystic/macrocystic radiologic pattern and the history of non-pancreatic malignancies were found to be significant predictors of tumour growth. Nevertheless, the estimated mean growth rate of SCNs for such levels was 0.34 cm/year, again slower than previously thought. Furthermore, tumour growth was found to increase with age (which was considered as a random effect); while—importantly—tumour diameter at the time of diagnosis was not a significant factor in the model. In particular, 33 of 145 patients had a tumour size >40 mm, and only six of them underwent a resection in the study period because of an increase in tumour size (but a significant fixed effect was always concurrently present).
The reasons for a more rapid growth in the macrocystic variants of SCNs and in patients with a history of other non-pancreatic malignant neoplasms remain unknown. The etiopathogenesis of the macrocystic lesions—in particular—is unclear; this subgroup of SCNs has been thought to be a particular form of microcystic lesions in which one cyst develops at the expense of the others. A hormone-dependent hypothesis has not been substantiated, and the proliferative index was found to be very low.15–17 In our series, the oligocystic/macrocystic variant accounted for 14.5% (a slightly greater incidence than in other reports)7 13 18 and was also associated with a greater resection rate in comparison with the microcystic counterpart (47.6% vs 7.5%, p<0.0001 and OR=11.21).
The association between pancreatic cystic neoplasms and a greater incidence of extrapancreatic neoplasms malignant and benign (from 16.8% to 52.0%) has been recently retrospectively demonstrated in patients with IPMNs.19 Most of these extrapancreatic neoplasms precede the diagnosis of the pancreatic cystic neoplasms, and a careful systemic check-up has been recommended in preoperative screening and during follow-up.20–24 However, a recent prospective study comparing the incidence of extrapancreatic cancers in patients with IPMNs with the expected incidence of the age- and sex-matched general population (median follow-up of 50 months) showed that IPMNs were not associated with systemic carcinogenesis except for pancreatic cancer.25 The incidence of previous extrapancreatic neoplasms in patients diagnosed as having SCNs (11.0%) was less than that reported for IPMNs, and no other malignancy (either extrapancreatic or pancreatic) was seen during the follow-up period. Although prospective studies in this field are lacking, our data suggest that the concept of a systemic check-up for SCNs may not be mandatory.
In summary, this study shows that estimated mean growth of asymptomatic or minimally symptomatic SCNs is slower than previously thought, especially in the first 7 years from the baseline evaluation. All SCNs are virtually benign, and lesions with a confident diagnosis of SCN can be initially addressed with a surveillance protocol. Patients presenting with a significant predictor of growth (oligocystic/macrocystic morphology and/or a history of non-pancreatic malignancies) are more likely to require a resection in the long term and should be informed in this regard. An operative approach may be also considered in case of a substantial deviation from the growth pattern shown here. In any case, tumour diameter at the time of diagnosis should not be used for decisional purposes.
On a practical-oriented basis, in the presence of an asymptomatic lesion in the head of the pancreas suspected for SCN and devoid of clinical concern (especially in the frail or elderly patient), a conservative management seems to be suitable, given the high risk of a pancreaticoduodenectomy. By contrast, the increasing use of minimally invasive approaches for distal pancreatectomy led to a more aggressive policy for most cystic neoplasms involving the body–tail of the pancreas. Nonetheless, further diagnostic evaluation including EUS with fine needle aspiration cytology and analysis of cyst fluid (especially in young to middle-aged women) might be worthwhile, and follow-up with non-invasive imaging seems appropriate if a confident diagnosis of an SCN is reached. The optimal interval between follow-up imaging tests in pancreatic cystic lesions is yet unclear. Although consensus guidelines recommend imaging on a semiannual or annual basis,26 recent studies have shown that cystic lesions presumed to be benign can be safely observed on a 2-year basis.27 According to the results shown here, a follow-up interval of 2 years may be optimal, although a verification of the present model and of the cumulative effects of potential predictors of tumour growth is required to establish guidelines with an greater level of evidence.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Download Supplementary Data (PDF) - Manuscript file of format pdf
The results of this paper have been presented in partial form at the 45th Annual Pancreas Club Meeting, Chicago, Illinois, USA, May 6–7, 2011.
Competing interests None.
Ethics approval This study was approved by Institutional Review Board.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.