Prospective evaluation of oral mucositis in patients receiving myeloablative conditioning regimens and haemopoietic progenitor rescue


Dr G. C. Jayson, Cancer Research Campaign Department of Medical Oncology, Christie Hospital NHS Trust, Wilmslow Road, Withington, Manchester M20 4BX, UK.. E-mail:


Four hundred and twenty-nine patients received myeloablative chemotherapy for solid and haematological malignancies in a bone marrow transplantation unit. Regimens appropriate to the tumour type were administered and haemopoietic reconstitution was achieved with peripheral blood progenitor cells (PBPC; n = 275), autologous bone marrow (auto-BMT; n = 69) or allogeneic bone marrow (allo-BMT; n = 85). World Health Organization (WHO) oral mucositis scores were collected prospectively from the start of chemotherapy (d 1) until d 28 or discharge. Oral mucositis (OM) was experienced by 425 (99%) patients and in 289 (67·4%) this was grade III or IV. Strong opiate analgesia was prescribed for a median of 6 d to 47% of patients. Univariate analysis suggested that the area under the OM curve (AUC; sum of daily mucositis grades, d 1–28) was associated with the myeloablative regimen, haemopoietic progenitor source (PBPC > allo-BMT > auto-BMT), use of myeloid growth factors and age. Multivariate analysis showed that the only independent risk factor for mucositis was the conditioning regimen (P < 0·00005). The mean OM AUC for high-dose melphalan (HDM) regimens (52 grade–days) exceeded busulphan (41), busulphan–cyclophosphamide (35), cyclophosphamide–total body irradiation (TBI) (34), cyclophosphamide–carmustine (BCNU) (20) and cyclophosphamide–etoposide–carmustine (CVB) (19). HDM regimens resulted in the highest mean peak OM (3·6), followed by busulphan regimens (2·6), cyclophosphamide/TBI (2·3) and cyclophosphamide–carmustine and CVB (1·4). Busulphan produced significantly delayed OM (median 3 d; P < 0·00005). There was a linear association between the area under the OM curve for each treatment group and the time to reach grade 3 OM (P < 0·00005), but no association with the time to reach grade 4 neutropenia (P = 0·24) or thrombocytopenia (P = 0·73), implying that haematological and mucosal toxicity are not associated. The cytotoxic regimen is the most significant determinant of OM. Studies investigating agents to ameliorate mucosal toxicity should be stratified according to cytotoxic regimen.

The principal toxicities of chemotherapy occur because of damage to rapidly proliferating cell populations, such as those in the bone marrow and the gastrointestinal tract mucosa. Oral mucositis (OM) has been recorded during the treatment of 39% and 67% of patients with non-head and neck cancer and leukaemia respectively ( Sonis et al, 1978 ; Raemaekers et al, 1989 ). Despite the introduction of the myeloid growth factors that ameliorated haematological toxicity, chemotherapy-associated mucositis remains a significant problem and there is no clear evidence to show that these growth factors improve mucositis ( Gabrilove et al, 1988 ; Welte et al, 1990 ; Pettengell et al, 1992 ; Schuchter et al, 1992 ; Gordon et al, 1994 ; Jones et al, 1995 ; Jones et al, 1996 ; Karthaus et al, 1998 ).

The most severe OM is encountered by patients who receive the most intensive treatments, and in patients receiving myeloablative chemotherapy the prevalence is between 28·6% and 100% ( Seto et al, 1985 ; Berkowitz et al, 1987 ; Sonis & Clark, 1991 ; Schubert et al, 1992 ; Woo et al, 1993 ). In this situation, oral ulceration typically lasts for a median of 6 d, with resolution of ulcers in most patients by 15 d ( Woo et al, 1993 ) and mucositis by 22 d after bone marrow transplantation (BMT) ( Garfunkel et al, 1994 ).

Oral mucositis is an important clinical problem because of the pain, the requirement for parenteral nutrition and the risk of mucosal infection and subsequent septicaemia. In many patients undergoing myeloablative therapy, it is the recovery of the oral mucosa, rather than haematological function, that delays the patient's discharge. New treatments are needed to reduce the duration and severity of mucositis, but these can only be developed once the natural history of mucositis has been described. Here, we report the clinical progress and multivariate analysis of the causes of oral mucositis in patients undergoing myeloablative therapy in a dedicated bone marrow transplantation unit.


Patients. Four hundred and twenty-nine patients underwent myeloablative conditioning and haemopoietic progenitor cell reinfusion in the Department of Haematological Oncology at the Christie Hospital between 1985 and 1997. Daily oral assessment was performed and mucositis was scored by the doctors and nurses in the transplantation unit according to the World Health Organization (WHO) mucositis scoring system ( Miller et al, 1981 ): grade I, erythema; grade II, painful ulceration, can eat; grade III, painful ulceration, cannot eat; grade IV, painful ulceration, requiring parenteral support or opiate analgesics.

If a mucositis score was missing for a particular day, the mean of the preceding and subsequent days' scores was used. In addition, if a patient was discharged with a mucositis score that was not 0, then the level ascribed for the remaining days (up to 28) was half the grade at discharge. However, no patient was discharged with grade 3 mucositis. For each patient, the initial disease diagnosis, the conditioning regimen, the haemopoietic progenitor cells infused, the use of haemopoietic growth factors, the duration of grades 3 and 4 neutropenia, the duration of grades 3 and 4 thrombocytopenia, the daily WHO OM score from the start of the conditioning regimen (d 1) for 28 d and the duration of opiate analgesia for OM were recorded prospectively.

Supportive care. Patients were managed in a purpose built leukaemia and bone marrow transplantation unit. Standard prophylaxis over the study period included: (i) chlorhexidine mouthwash solution used three times daily from admission to discharge; (ii) acyclovir 200 mg five times daily from d −3 to +6 weeks; (iii) oral imidazole antifungal; (iv) cotrimoxazole 960 mg twice daily (bd). Parenteral nutrition was not used as a prophylactic measure, but was prescribed when oral nutrition was not sustained. Analgesia was prescribed for mucositis according to the subjective report of pain by the patient. The WHO ladder of analgesia was used so that patients received simple analgesics initially. If more severe pain occurred or if the analgesia was insufficient, simple opiate-based analgesics were initially prescribed and then morphine-based opiates that were given parenterally if the oral route of administration was not possible. Graft-versus-host disease prophylaxis for those patients undergoing allogeneic transplantation was with cyclosporin A. Some patients who were undergoing allogeneic transplantation also received methotrexate. The platelet transfusion threshold was 20 × 109/l. Haemopoietic growth factors were routinely used in the 1990s and administered daily from day 1 or day 4 after haemopoietic progenitor infusion until the neutrophil count exceeded 1·0 × 109/l.

Broad spectrum antibiotics were commenced empirically if the patient's temperature exceeded 38°C or if it was greater than 37·5°C with signs of infection. Empirical intravenous antifungal therapy was commenced if pyrexia persisted for 72 h.

Myeloablative regimens.Table I shows the myeloablative regimens.

Table I.  Myeloablative therapy regimens .
RegimenDosesFrequencyDaysDiagnostic groups
  1. Cy, cyclophosphamide; Bu, busulphan; VP-16, etoposide; BCNU, carmustine; HDM, melphalan; TBI, total body irradiation 200 cGy per fraction twice daily for 3 d (total dose 12 Gy); ALL, acute lymphoblastic leukaemia; ANLL, acute non-lymphoblastic leukaemia; CML, chronic myeloid leukaemia; NHL, non-Hodgkin's lymphoma; HL, Hodgkin's lymphoma; MM, multiple myeloma. od, once a day; bd, twice a day; qds, four times a day.

Cy60 mg/kgod1, 2ALL, ANLL, CML
TBI200 cGybd3–5 
Bu1 mg/kgqds1–4NHL, ALL, ANLL
Cy50 mg/kgod5–8 
Bu1 mg/kgqds1–4ANLL
Cy1·5 g/m2od1–4HL
BCNU600 mg/m2od5 
Cy1·5 g/m2od1–4HL
VP-16200 mg/m2od1–4 
BCNU300 mg/m2od5 
HDM110 mg/m2od1MM
TBI200 cGybd3–5 
HDM200 mg/m2od1MM

Statistics. The severity of OM was measured in two ways: (i) the sum of the OM grade for each day, days 1–28 (where d 1 was the start of the conditioning regimen); this can be thought of as the area under the OM curve (minimum possible = 0 and maximum possible = 112.); (ii) the time, in days, to grade 3 OM from d 1 (as defined above). For cases in which grade 3 OM was not reached, the time was censored at d 28.

Two multivariate analyses were carried out: (i) multiple regression on the area under the OM curve; (ii) Cox regression on the time to onset of grade 3 OM. The censored variable being days to grade 3 OM.

The explanatory variables considered in each case were the conditioning regimen, haemopoietic progenitor type [allogeneic BMT/autologous BMT/peripheral blood progenitor cells (PBPC)], the growth factor (used/not used), the age of the patient and the sex of the patient.

Where appropriate, the data were also analysed using a one-way analysis of variance ( anova) followed by Duncan's multiple range test. For the time to grade 3 OM, the log rank test was used. To consider whether the area under the curve was associated with time to grade 3 OM, the area under the curve of each patient was categorized into one of seven groups. These groups were chosen to give equally spaced area bands across the middle five of the seven groups and reasonably similar numbers in these bands: < 15, 16–25, 26–35, 36–45, 46–55, 56–65 and 65+ (grade–days).


Four hundred and twenty-nine patients underwent myeloablative conditioning ( Table II). Eight patients were excluded because of incomplete data. Forty-five patients were excluded from further analysis because of small numbers in each of the individual conditioning regimens, and 15 patients were excluded because of death within 28 d of commencing treatment (four fulfilled both criteria). Thus, 365 were available for analysis, of whom 67 received allogeneic bone marrow (allo-BMT), 62 patients received autologous marrow (auto-BMT) and 236 patients received peripheral blood progenitor cell reinfusion (PBPC). Ninety-three patients received no growth factor, whereas 248 patients were treated with granulocyte colony-stimulating factor (G-CSF) and 24 patients with granulocyte–macrophage colony-stimulating factor (GM-CSF).

Table II.  Patient characteristics.
 Frequency of all
(n = 429)
Percentage of all
(n = 429)
Frequency in group for analysis
(n = 365)
Percentage in group for analysis
(n = 365)
  1. The table shows the statistics for the whole group and also for the 365 patients available for analysis, when the less commonly used regimens had been excluded.

  2. BEAM, BCNU, etoposide, Ara-C, melphalan; Carbo, carboplatin.

 Mean (SD, range)36·7 (13·3, 14–68) 37·2 (13·5, 14–68) 
Treatment Regimen
Haemopoietic cell type
Growth factor

Oral mucositis was experienced by 362 (99%) patients and 246 (67·4%) patients encountered at least grade 3 toxicity. Strong opiate analgesia was prescribed for 173 patients (47%) for a median of 6 d. The area under the mucositis curve for the 408 patients, for whom 28 d of data were available, was 37·9 grade–days [standard error (SE) 0·87], whereas that of the 365 patients who were entered into the rest of the analysis was 37·4 grade–days (SE 0·897), suggesting that the exclusion of infrequent regimens from the analysis did not significantly affect the results.

Mucositis induced by the seven conditioning regimens

The data showed that the regimens could be distinguished according to the maximum grade of mucositis, the time of onset of maximum mucositis and the rate of onset and recovery ( Fig 1).

Figure 1.

Mean mucositis score associated with each treatment regimen (SEM ± 0·5 grade).

The regimens incurring the most severe oral mucositis were those that incorporated high-dose melphalan [HDM and HDM/total body irradiation (TBI): mean AUC mucositis score of 52 grade–days]. Between 12 and 14 d after melphalan administration, these regimens were associated with a mean maximum WHO mucositis score of 3·6. The development and resolution of mucositis occurred at a similar rate for both regimens, although the HDM/TBI curve lagged 1 d behind the HDM curve.

Cyclophosphamide/TBI (CyTBI) (mean AUC 34 grade–days), busulphan (mean AUC 41 grade–days) and busulphan–cyclophosphamide (mean AUC 35 grade–days) were associated with the same mean maximum level of mucositis (2·6). However, busulphan-containing regimens were associated with a 3-d delay in the onset of mucositis (P < 0·00005, one-way anova).

The regimens that were least toxic to the mucosa were cyclophosphamide–carmustine (CyBCNU) (mean AUC 20 grade–days) and cyclophosphamide–etoposide–carmustine (CVB) (mean AUC 19 grade–days) that were associated with gradual onset mucositis that reached a mean peak WHO score of 1·4.

Statistical analysis

The multiple regression analysis was performed to identify the most significant factors associated with oral mucositis with respect to the ‘area’ under the OM curve. Univariate analysis showed that the conditioning regimen, the haemopoietic cell source, the use of myeloid growth factors and age were associated with the area under the curve of mucositis. The conditioning regimen was the most significant factor (P < 0·00005).

Using a forward stepwise approach in the multiple regression, only the conditioning regimen retained statistical significance. The univariate models are also shown in Table III.

Table III.  Associations with area under the curve of mucositis.
Univariate variablesRegression coefficient βSE(B)Univariate PWith treatment group in
model P
  1. Multiple regression uni- and multivariate analyisis of factors associated with area under the curve of mucositis.

Treatment group
 Constant18·772·698< 0·0005
 HDM TBI33·753·179  
 Constant24·412·528< 0·00050·299
HPC source
 Constant29·982·137= 0·0010·194
Growth factor
 Constant32·341·752= 0·0010·071
 Yes 6·792·029  
 Constant37·231·12= 0·800·435

We found a close association between the area under the curve of mucositis and the time to the development of WHO grade 3 mucositis for each patient (P < 0·00005, log rank), and Table IV shows the Cox analysis of the same variables using this new end-point.

Table IV.  Associations with time to development of grade 3 mucositis.
Univariate variablesRegression coefficient βSE(B)Univariate PWith treatment group in
model P
  1. Cox regression uni- and multivariate analysis of factors associated with time to development of grade III mucositis.

Treatment group
 CyTBI1·3530·4643< 0·000005
 HDM TBI3·0060·4687  
Age0·02060·0051= 0·000040·12
HPC source
 Allo 0·30180·2304= 0·00190·36
 PBPC0·598 0·1906  
Growth factor
 No0·0000Baseline= 0·00670·32
 Male0·0000Baseline= 0·800·35

Both sets of analyses revealed the same results ( Tables III and IV). The treatment group was the most significant predictive factor for oral mucositis. Although increasing age, progenitor cell source and use of myeloid growth factors were significant on univariate analysis, the significance of these factors was lost when treatment group was taken into account (final column of each table). With reference to the haemopoietic progenitor cell source, the data show that the least mucositis was observed in patients undergoing auto-BMT, whereas patients undergoing allo-BMT and PBPC transplantation experienced more marked mucositis.

Myeloid growth factors

The data suggest that for CyTBI, CyBCNU and the melphalan regimens, those patients who were treated with myeloid growth factors appeared to experience greater mucositis. Although the univariate analysis highlighted this result, the statistical significance of the result was lost when treatment group was taken into account whether we considered area under the mucositis curve or time to the development of grade 3 mucositis.

Haematological and mucosal toxicity

We wanted to evaluate the relationship between haematological and mucosal toxicity to determine whether the regimens that were toxic to haematological stem cells were also toxic to mucosal stem cells. The hypothesis was that those patients who had the most rapid onset neutropenia or thrombocytopenia would show an ordered relationship with time to onset of grade 3 OM. For this part of the analysis we focused only on patients who received PBPC (n = 236) and, as intended, all patients incurred grade 4 haematological toxicity. There were differences in the interval between the start of treatment and the onset of grade 4 neutropenia (P < 0·0005, one-way anova) and grade 4 thrombocytopenia (P < 0·0005, one-way anova) according to the different regimens. However, there was no relationship between the interval to the onset of either haematological toxicity and the time to onset of grade 3 mucositis (neutropenia, P = 0·24; thrombocytopenia, P = 0·73; log rank test).

Disease and treatment

We analysed the area under the mucositis curve for patients treated with busulphan and cyclophosphamide for leukaemia (n = l6) and lymphoma (n = 68). The patients with lymphoma incurred a mean of 35·1 grade–days, whereas those with leukaemia had 34·7 grade–days of mucositis (t-test, P = 0·92), suggesting that the regimens rather than the different disease types were responsible for mucositis.


The development of peripheral blood progenitor haemopoietic reconstitution significantly ameliorated the neutropenia associated with myeloablative therapy and revealed epithelial toxicity, such as mucositis, as the principal factor that limited further dose escalation. High doses of chemotherapy impede mucosal integrity, resulting in pain, increased risk of infection and an impaired nutritional state ( Ruescher et al, 1998 ; Rapoport et al, 1999 ). New approaches to reduce mucositis are required, but their effective introduction requires a large data base of knowledge upon which these treatments can be developed. This study is the largest prospective analysis of mucositis in patients undergoing myeloablative conditioning regimens.

Mucositis was common in our population despite the use of prophylactic antimicrobial medication. The prevalence was 99%, with approximately 70% of patients incurring grade 3 toxicity and 47% requiring strong opiate analgesia, statistics that are high when compared with other series ( Seto et al, 1985 ; Berkowitz et al, 1987 ; Woo et al, 1990 , 1993; Sonis & Clark, 1991; Schubert et al, 1992 ; Chou et al, 1996 ). Our data support the concept that melphalan-containing regimens are associated with the most severe mucositis ( Lazarus et al, 1983 ; Ayash et al, 1994 ; Cunningham et al, 1994 ; McCowage et al, 1995 ). Other authors have suggested that total body irradiation is the most toxic agent ( Schubert et al, 1992 ; Rapoport et al, 1999 ) and this difference may be due to differences in fractionation regimens.

The relative time of onset of mucositis according to the regimens used has not been reported before. It is striking that the onset and recovery of busulphan-associated mucositis was delayed, whereas the onset of melphalan-associated mucositis was apparent approximately 1 week earlier. This is most likely to be a result of the difference in schedule for the two agents, although an alternative hypothesis is that the two drugs affect different stem cell populations. However, the fact that both the haematological and mucosal toxicities of busulphan were delayed suggests that the drug schedule is responsible rather than a difference in stem cell susceptibility to the agents.

The univariate analysis identified age, haemopoietic progenitor source, use of myeloid growth factors and drug regimen as risk factors for mucositis, but following multivariate analysis only the conditioning regimen retained independent predictive power. This was because there was an association between disease type, conditioning regimen and haemopoietic source. The prevalence of grade 3 mucositis was greater in the patients receiving PBPC (72%) than in those receiving allo-BMT (65%), contrasting with the data of Rapoport et al (1999) , yet nearly all of our patients who were treated with melphalan received PBPC reinfusion. We cannot exclude that the disease type contributed to the mucositis as there was a close association between disease and treatment regimens. Although it is likely that the drug regimen is responsible for the mucositis, the amount of previous treatment and paresis of immunoglobulin function are closely related to each disease and may have had an impact on mucositis. Nevertheless, we compared the mucositis of patients treated with busulphan and cyclophosphamide for leukaemia and lymphoma and found no statistical difference, implying that the drug regimen, rather than the underlying disease, is the main risk factor for mucositis.

The data revealed that patients receiving myeloid growth factors incurred greater mucositis. Although some reports have suggested that these cytokines afford some mucosal protection ( Gabrilove et al, 1988 ; Gianni et al, 1989 ; Herrmann et al, 1990 ), others have not ( Meropol et al, 1992 ; Pettengell et al, 1992 ; Schuchter et al, 1992 ; Jones et al, 1995 ; Legros et al, 1997 ), and in this case the excess mucositis is probably due to the use of G- and GM-CSF in patients treated with busulphan or melphalan.

Recent work has suggested that the duration of severe neutropenia was a risk factor for mucositis ( Rapoport et al, 1999 ). However, it is of interest that we did not find any relationship between myelosuppression and mucositis even on univariate analysis.

These data have arisen from the largest prospectively recorded evaluation of mucositis in patients undergoing myeloablative therapy. The results show that the conditioning regimen is the principal determinant of mucositis and that melphalan-containing regimens are the most toxic to the mucosa. The data should serve as a baseline upon which agents designed to protect the mucosa, such as interleukin (IL)-11 and keratinocyte growth factor (KGF), can be introduced. The close correlation between interval to the onset of grade 3 mucositis and area under the mucositis curve identifies the former as a potential new end-point in mucositis research and this should be prospectively evaluated in further clinical trials.


We would like to thank the nurses and junior doctors on the adult leukaemia unit at Christie Hospital for their help in acquiring this data set.