Prognostic factors in patients with non-- Hodgkin ' s lymphoma presenting with gastrointestinal tract symptoms

BH WEINERMAN, B MACDOUGALL, l CARR. Prognostic factors in patients with non-Hodgkin's lymphoma presenting with gastrointestinal tract symptoms. Can J Gastroenterol 1991;5(1):5-10. Forty-one individuals who presented with symptoms referable to the gastrointestinal t ract were extracted from the authors' total review of non-Hodgkin's lymphoma in Manitoba from 1968 through 1978. O nly cases at stages I to Ill were included, and there were 22 males and 19 females. Sixteen non-Hodgkin's lymphomas arose in the small bowel, LS in the stomach and 10 in the colon. The natural history of this group of gastrointestinal non-Hodgkin's disease is described, and univariate and muluvariate analyses were done using the variables of sex, pathology, stage, resection, site of disease and init ial chemotherapy. The median survival of the group was 28 months. Sex and stage appeared to be important prognoscically, but after multivariate analysis, only the female sex appear to be a good prognostic variable. There was a suggestion that resection should be attempted in these lesions, but there was not a large sample size in this group. In add it ion, it was felt that this group of individuals followed the same survival pattern as did poor prognosis non-Hodgkin's lymphomas of nongastrointestinal origin.

I T IS GENERALLY AGREED TIIAT lymphomatous involvement of the gastrointesti nal tract is the most common site of extranoJal primary non-Hodgkin's lymph oma invo lvement (1,2 ).Several d ifferent prognostic factors have been reported in mu ltiple papers, often with d iffe ring conclusions (3 ).Even histological suh types , generally the most important prognostic factor in systemic disease ( 4 ), were found, by some, to be less important than stage or othe r facto rs such as lactate dchydrogenase, sex and perforation.One of the questions that arises is whethe r patie n ts presenting with symptoms referable to the gastrointestinal tract have a d iffere nt natural history from those with ocher lymphomas of the same stage anJ histological type, and whether, in this group, the re are any prognostic variables that may be helpfu l to the clin ician.The studies that have examined t hese questions (2,3,(5)(6)(7) have not been based on an entire population experience (such as the Manicoba T umou r Registry) o r have not been solely confined to people presenting with symptoms.
1968-78 ( 4), those individuals who presented with symptoms referable to the gastrointestinal tract.Symptoms and signs were often not clearly described in the charts, but anyone who presented with an ab<lominal mass thought co be of gastrointestinal origin, abdominal pain of gastrointestinal origin, or altered bowel function as a presenting complaint, was consi<lered symptomatic.Weight loss by itself was not considered a gastrointestinal symptom.Two patients of the 41 were said to have a history of sprue, not confirmed on biopsy.Material at the time of diagnosis of lymphoma was not re-exa mined for this condition.Sixteen ocher patients had examinations of their bowels noted in this review, none of which were noted to be sprue.One patient had benign peptic ulcer, one inflamed gascnc mucosa, an<l one a prominent plasma cell infiltrate.All others were negative.
Cases of gastrointestina l non- Hodgkin's lymphoma were examined in order to describe their natural history, and to elucidate the survival implications of sex, pathology, stage, resection, site of disease and initial c hemotherapy.

PATIENTS AND METHODS
The survival of all a<lult patients with non-Hodgkin's lymphoma from the Mani coba Tumour Registry has been previously reported ( 4 ).The authors extracted from this database (473 cases -all adult, age greater than 18 years) individuals who were confirmed by biopsy co have gastrointestinal non-Hodgkin's lymphoma and who presented with gastrointestinal complaints.All pathology slides were reviewed by one of the authors (IC) and, if slides were not available, the case was not include<l in the analysis (five cases).The cases were classified according to the Rappaport classification (8) and the International Working Formulation (9)

RESULTS
Forty-one patients from t he total review of 473 non-Hodgkin's lymphomas (1968-77) who presented with gastrointest in al sym ptoms anJ had stages 1 to Ill primary gastrointestinal lymphomas were ident ified.A ll pathology slides were reviewed.S ixteen non-Hodgk in's lymphomas arose in the small bowel (39%), 15 in the stomach {37%), and 10 in the colon (24%).T he ages of the patients ranged from 18 to 90years with a median of 61 and a mean of 59.3.There were 22 males ( includ ing one with T umer's syndrome who was not analyzed when examining the effect of sex) and 19 females.Seveney-three per cent of patients presented with abdominal pain, 29% with significant weight loss and 41 % with constipation or some combi natio n of dia rrhea and constipation .An abdom ina l mass was noted on ph ysical exa minat io n in 22% (Table 1).
The median survival of the entire group was 28 mon ths (95% confidence interval 22 to 34) (Figure 1).Patients were separated into good and poor prognosis groups based o n the Rappaport dassificauon .Figure 2 shows the actuarial surv ival plots of those groups compared co the entire tumo ur registry population from the same period .There is a suggestion of better survival and a plateau in the good prognosis group, but the numbers are coo small to make any firm statemen t (Wilcoxon's statistics, P=O.l).There is no difference in survival between t he two poor prognosis groups (gastrointesti na l and nongastrointestinal), anJ they fo llow similar sur-v1Val curves, with med ian survivals of 18 months in the gastrointestinal group and 17 months in the entire group of lymphomas.
When surviva l curves were plot ted according to stage, the re were 16 stage I patients with a med ian survival of 6 1 months, 19 stage 11 with a median survival of 12 months, and six stage 111 {one who was actua lly stage IV by virtue of a single liver nodul e on laparoromy) with a median surviva l of n ine months.There was a suggestive difference in survival between stage l and stages II and Ill, but this d id not reach statistical significance (0.05<P<O.l ) (Figure 3 ).When the lymphomas were analyzed using sex as t he variab le there was a sign ifican t diffe rence in surviva l.Females had a median survival of 66 months and an apparent plateau at that leve l , while males experienced a median surviva l of 12 months (P=0.02)(Figure 4).Four of t he six 'good prognosis' patien ts were in t he female group, as were 10 of the 16 stage I patients.Extracting t he 'good prognosis' patien ts from both groups did not materially change the surviva l curves, although the median su rvival for t he female group was at ta ined at 60 rather than 66 mo n ths .Statist ica l significa n ce remained at P=0.02.Age was not a facto r, with most of the patients being in the o lder age group.
Seventeen of 41 patients received rad io th erapy eith er combineJ with surgery, or alone as primary treatment.T h is d id not appear to be a significanr factor in surviva l, but the number of cases were too few to make a statement.
There was no J ifference in surv ival in patients who were pri marily resected (P=O.l ) (Figure 5), a lthough o n ly 12 were not resected and hence the sample was small.Seven of the 41 patients deve loped perforation soon after in itial presentation : fo ur of 12 in the nonsurgical group, and three of 29 in the surgicall y resected group.Three of the seven perforated patients died with sep-  2).The site of lymphoma was also examined, anJ this Jid not correlate with survival (P=0.8)(Figure 6).Initial chemotherapy versus no initial chemotherapy was also examined.There was, in fact, a tendency fo r the group receiv- ing initial chemotherapy to do worse, with a P value between 0.05 and 0.1 (Figure 7).The six prognostic variables used were a ll compared in a univariate fashion using the log rank test ( 11 ).The variables which seemed to show some prognostic importance (sex, stage and initial chemotherapy) were considered in the proportiona l hazards regression model (12).By this method only female sex appeared to be an inJependent vanable in prognosis.

DISCUSSION
The authors rev iewed 41 non-Hodgkin's lymphoma patie nts extracted from a tumour registry, who presented with gastrointestinal complaints.They restricted the analysis to patients who had stages l to l ll non-Hodgkin's lymphoma.In univari ate analysis only female sex, lesser disease stage, and resection of the primary were important m predicting improved survival.In mu ltivariate analysis only female sex was important ( 12).
Primary lymphomas of the gastro• intestinal tract are re latively rare (2, 13 ), although the gastrointestinal tract is frequently involveJ as a secondary site in widespread lymphomas (I).Many articles have identified important prognostic variables on review of their cases, including clinical stage, age, sex, initial presentations with perforations, erythrocyte sedimentation rate, tumou r size, primary gastrointestinal site, histological grade {1 -3,5,7,8) and the addition of radiotherapy (6) or chemotherapy (14 ).The authors wondered if by restricting cases to tn• dividuals who clearly presented wi th gastrointestinal symptoms, a more specific pattern would emerge.Unfortunately, the resulting sample size was small , and nothing emerged as significant other than female sex.
A ozasa et al {15, 16) reported a poorer prognosis fo r males, but others have not reported th is (3).This poorer prognosis was a definite pattern in the present results which cannot be ex• plained by any other variable.Some authors have advocated resection of the primary before definitive therapy to avoid possible perforarion (2).A high percentage of the present nonresected cases perforate<l ( three of 12), but this did not reach statistical significance, and the selection that went into the choice of therapy is unknown.Nevertheless, it is the authors' impression that resection should be strongly recommended in this group of patients, since 25% of the nonresected group perforateJ compared to 10% of the resected group.
The finding thal chemotherapy may have an adverse effect on survival was not borne out by multifactorial analysis.Others have founJ a defi nite benefit from this modality of therapy (l, l 7) .The lack of benefit might be explained by the time period of the study, in which: more advanced multiagent anthracycline-based chemotherapy was not used; there may have been a selection bias in choosing patients to receive chemotherapy; and there may have been a selection bias in the criteria used to include patients as receiving chemotherapy.The authors Jid not make any attempt to decide whether o r not therapy was adequate, and any patient who received an y therapy, whether single or multiple agent within the first two months, was accepted as having received chemotherapy.
The distribution of lymphomas was different in this se ries compared to other large series such as that of Azab et al (3) and Lewin et al (5 ).Thirty-seven per cent of the prese nL cases were stomach, compared to 36% and 56% in the other series, respectively; 39% small bowel.compared to 34% and 36% in the other two series, respectively; and iinally, 24% large bowel compared to 7% and 9%, respectively.However, in these other cited series, ileocecal lymphomas were classified separately.Four of the present large bowel lymphomas were cecal, and if one removes these from the present group, large bowel lymphomas represent 1 7% of the group, probably not very different from what ochers report.
In summary gastrointestinal non-Hodgkin's lymphoma does not appear to have a particular set of symptoms that would alert a clinician to the possible diagnosis.T he authors' analysis followed the same survival curve as did  --'-----'-~'----'-----'-~'----'--'-~'----'--'-~ their review of all lymphomas in the same time period, but a few differences seem noteworthy.Female sex is an independent predictor of survival, which was not found in the authors' analysis of the entire group of non-Hodgkin 's lymphoma (4).Stage was also not found to be a factor, and although this has been fo und by others (2, 15 ), the authors may have influenceJ this by excluding stage IV patients.They did not have a large group of 'good prognosis' patients, so histology as a facto r could not be properly assesseJ.Most importantly, the same type of survival curve was seen with this group as with the en tire group, so that it would seem reasonable to apply the same principles of therapy to this group of lymphomas as are used for no ngastro in testi n al non-Hodgkin 's lymphoma.Resection, if possible, may still be indicated , although perforation can occur in resected individuals, from unrecognized areas.

Figure 2 ) 3 Figure 3 )
Figure 2) Actuarial sur11111al plor,s of good versus poor prognosis groups based on the Rappaport classificarion

Figure 4 )Figure 5 )
Figure 4 ) Aciuanal survival plots of males versus females with gastrointestinal lymphomas