This Article Abstract Figures Only Full Text (PDF) Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Dayes, I S Articles by Abuzallouf, S Search for Related Content PubMed PubMed Citation Articles by Dayes, I S Articles by Abuzallouf, S

Local tumour control in women with carcinoma of the cervix treated with the addition of nitroimidazole agents to radiotherapy: a meta-analysis

¹ Department of Medicine, McMaster University, Hamilton, Ontario, Canada, ² Department of Radiation Oncology, Juravinski Cancer Centre, Hamilton, Ontario, Canada and ³ Department of Radiation Oncology, Kuwait Cancer Control, Kuwait City, Kuwait

Correspondence: Dr Ian Dayes, Juravinski Regional Cancer Centre, 699 Concession St., Hamilton ON, L8V 5C2, Canada

	Abstract

Top Abstract Introduction Methods and materials Results Discussion Conclusion References

 
The purpose of this paper was to estimate the effect of nitroimidazoles on the local control and overall survival of women receiving radiotherapy for carcinoma of the cervix. Sources searched included Medline, Cancerlit and national cancer organizations. Proceedings of meetings were hand-searched. Trial selection and quality score were performed in duplicate. Data extraction was performed by a single author. Five trials involving 849 patients were included. Median follow-up was typically 4 years or greater. The odds ratio (OR) for local recurrence did not demonstrate a significant effect (OR: 1.14; 95% confidence interval (CI): 0.78–1.66). The difference in mortality was also non-significant (OR: 1.26; 95% CI: 0.95–1.66). A significant increase in neuropathy was found (OR: 3.21; 95% CI: 1.36–7.55). Subgroup analysis did not reveal any sources of heterogeneity between trials. Despite five published randomized trials, evidence supporting the use of nitroimidazoles in the treatment of cervical cancer is lacking. Meta-analysis revealed no significant effect on local tumour control with a weak, non-significant trend suggesting a decrease in overall survival. There is, however, a significant increase in the rate of neurotoxicity with the use of these compounds. This overview can not support the use of these agents.

	Introduction

Top Abstract Introduction Methods and materials Results Discussion Conclusion References

 
It has long been known that hypoxic tissues, both normal and malignant, are resistant to the effects of radiation. Hypoxia in tumours has been hypothesized to be an important cause of treatment failure in a variety of human malignancies and as such, efforts have been made to overcome this problem, including the use of compounds that may mimic the action of oxygen in the radiated cell [1].

The best studied compounds are the nitroimidazoles. Data from in vitro studies suggest that these drugs can overcome hypoxia-induced radioresistance. Clinical studies, however, have been less promising. One limitation of these studies is a lack of statistical power, as many trials have failed to detect an effect of the addition of these drugs. As such, the following meta-analysis has been performed to study this issue further.

A previous meta-analysis has been reported on the role of radiosensitizers in multiple tumour sites [2]. This report included data from over 7000 patients from 50 trials in 6 tumour sites. Disease sites, including cervix, were examined together and individually. This review reported a statistically significant improvement of local tumour control and overall survival from the use of nitroimidazoles across all disease sites but found a non-significant trend against their use in cervix cancer. Unfortunately, little information was given regarding the search strategy. In addition, the statistics used were poorly described. It would appear that patient numbers and events were summated and odds ratios calculated without accounting for individual trial contributions. Summary measures for toxicity were not reported. The following meta-analysis was performed in an effort to overcome some of these methodological limitations. In addition, it was felt that longer follow up from previous trials and additional trials may have become available since publication of the previous report.

Local tumour control was chosen as the primary outcome because of the underlying principle guiding the use of radiosensitization. If such an agent were to affect the ability of radiotherapy to kill malignant cells, it may not have any impact on survival as patients could still succumb to metastatic disease. Thus, in addition to being an important clinical end point, local control was felt to provide the best measure of the biological rationale for the use of nitroimidazoles.

	Methods and materials

Top Abstract Introduction Methods and materials Results Discussion Conclusion References

 
Identification of relevant research
Relevant articles were identified using the following strategies. First, a search for reviews was performed on Medline using the following terms: cervix, review and one of either radiotherapy, radiosensitization or nitroimidazole. Articles were then reviewed for primary sources of data by examination of the reference sections. In addition, the "Related Articles" function was used on Medline for each relevant review.

Second, the electronic databases Medline and Cancerlit were searched using the following strategy. Briefly, the names of all nitroimidazoles listed in Medline as MESH headings were used as search headings along with "cervix cancer", "cervix neoplasms", "radiotherapy" or "treatment outcome" as major MESH headings and "randomized controlled trial" or "clinical trial" as publication type. The "Related Articles" function was used on Medline for each potentially relevant citation.

Online collections of reviews and guidelines were searched. These included reports by the Cochrane Collaboration, Cancer Care Ontario, British Columbia Cancer Agency, the American Society of Clinical Oncology, the American Society of Therapeutic Radiation Oncology and the American College of Obstetricians and Gynecologists.

Proceedings of annual meetings of the American Society of Therapeutic Radiation Oncology and the American Society of Clinical Oncology were hand-searched for the previous 10 years.

Finally, first authors of identified papers were sent e-mail messages asking if they were aware of any other studies in the field. In addition, authors were asked if any later publications of their papers had occurred or if any were planned. Colleagues were asked to identify any trials missed by the search strategy.

Selection of relevant research
Selection criteria were defined prior to execution of the above search strategy. All eligible studies were randomized controlled trials. The study population was patients with histologically confirmed, locally advanced carcinoma of the cervix treated with radical doses of radiotherapy (external beam and brachytherapy to a total dose of 60 Gy) with curative intent. Subjects were to be randomized to a concurrent systemic nitroimidazole group or no additional therapy/placebo. Outcomes must have included local disease control and disease-free survival and/or overall survival and/or toxicity. Trials were excluded if patients with recurrent disease were included or if patients received cytotoxic chemotherapy or surgery.

Potentially eligible papers were selected by the abstract, or title if the abstract was not immediately available. Once identified, full reports were retrieved. Two radiation oncologists evaluated the papers independently, with degree of agreement measured by an unweighted Cohen's kappa. To facilitate any ensuing discussions resulting from potential differences in selection, each oncologist identified and recorded the criterion which most identified each paper as being ineligible. Once each reviewer had selected reports, disagreements were resolved by consensus. If consensus was not reached, a third radiation oncologist was available to arbitrate decisions, if required.

Assessment for the completeness of the retrieval process and possible publication bias was performed by the use of a funnel plot by plotting effect size vs sample size [3].

Assessment of methodologic quality
Previous research has indicated that the inclusion of low quality clinical trials into meta-analyses may lead to an increased estimate of treatment benefit [4]. Therefore, following the selection of all reports, methodology was assessed by scoring against a checklist developed via minor modification of a previously reported assessment tool [5]. Certain elements of the original checklist were changed slightly to increase their relevance for radiation oncology trials and to help make instructions more explicit. Items in the checklist score for the quality of reporting randomization and its concealment, blinding to treatment group, outcomes, study population, treatments, follow up, toxicity and issues of statistical assessment and reporting. Assessment was performed in duplicate, with a possible maximum score of 15 for each study. As the relationship between true quality of a paper and its subsequent score is unknown (e.g. linear, asymptotic), scores were converted to ranks. Comparison between assessors was performed using Spearman's rank correlation coefficient. Studies with a mean score of seven or less were excluded.

Data extraction
A data extraction form was created prior to study selection with extraction performed by a single observer (ID). Extracted details of studies included paper identification (author, journal, year), sample size, the study population (including staging investigations and exclusion criteria) and treatment details of both radiation (external beam and brachytherapy doses) and nitroimidazole (compound, dose, route, frequency, duration). The primary outcome of interest was local control rate. Other outcomes included median follow up, proportion of enrolled patients not analysed, disease-free survival, complete response rate, overall survival and severe toxicity rates.

Data analysis
Outcomes were converted to odds ratios for the purpose of analysis. As odds ratios are based upon relative odds of experiencing an event, they were felt to be the most robust treatment effect measure as total events would likely be reported at different times (and with different proportions of patients at risk) between the studies. This assumes that relative event rates remain largely stable throughout the time of follow up (as is also true for the comparison of survival curves using hazard ratios). Unfortunately, this does not allow for censoring or time to event, a difficulty which has been raised in a recent meta-analysis regarding cervical cancer [6]. In addition, the odds ratio is ideal for the pooling of summary measures in a meta-analysis [7]. Pooling of data was performed using the random effects model [7], with results reported as pooled odds ratios (±95% confidence intervals).

Examination for sources of heterogeneity was performed by subgroup analysis. Factors selected for examination a priori included severity of cases (clinical stage of disease), type of nitroimidazole used, publication status and methodological rigour as assessed by the 15 point quality score.

	Results

Top Abstract Introduction Methods and materials Results Discussion Conclusion References

 
Identification of relevant research
During the initial search, 390 citations were felt to have potential relevance. Examination of titles and/or abstracts resulted in the rejection in the majority of these trials. The most common reasons for rejection included reporting of non-clinical studies, trials involving non-cervical cancers and reports on chemotherapy. 32 potentially relevant articles were retrieved following this initial search. There was agreement between the two reviewers on six articles for inclusion, 24 articles for exclusion and disagreement on two articles. The resulting unweighted Cohen's kappa was 0.82, suggesting good agreement. Disagreement occurred concerning two articles with respect to the nature of the radiotherapy. Upon further examination and discussion of the methods sections of the articles, both were accepted for inclusion. One of the trials was reported on several occasions. Thus, a total of five unique studies were included. All were published in full-length journal form. All five studies had been identified in the original Medline search. Four of these studies were also found in the subsequent Cancerlit search. One study was also found in abstract form by the hand search. No additional studies were identified by any of the subsequent search strategy.

Inspection of a funnel plot of sample size versus treatment effect was not found to be helpful with a total of only five data points. No attempt at statistical analysis was made as evidence suggests very low power in such analyses when fewer than 10 trials have been identified [8].

Validation of selected reports
Of a possible maximum of 15, scores ranged from 8 to 14 for one reviewer (ID) and 10 to 14 for the other (SA). Spearman's rank correlation coefficient was 0.88. Significance testing under the null hypothesis that the correlation is zero is not possible under these conditions (i.e. only five studies). A minimum of six ranks is required to make statistical significance (p<0.05) mathematically possible [9].

Summary of selected studies
A total of five clinical trials was identified as being eligible [10–15], with a total of 849 patients available for analysis (Table 1). Median follow-up was at least 4 years in all but one study. Two trials were reported as interim analyses but despite 17 and 8 years since publication, no further analysis was reported [10, 11]. The author was contacted and did not report any planned further analysis. One of these studies was closed early because of an excess of deaths in the treatment group [11]. A further trial was closed to accrual prior to reaching its planned sample size because an interim analysis suggested a very low likelihood of achieving results consistent with any clinical benefit [12].

View larger version (14K): [in this window] [in a new window]   Figure 1. Odds ratios for local control rates in nitroimidazole cervix trials. Left hand columns represent author (year of publication), number of patients, odds ratio (OR), lower and upper 95% confidence limits (95% CI). Odds ratios of less than one suggest improvement of local control rates with the addition of a nitroimidazole.

View larger version (14K): [in this window] [in a new window]   Figure 2. Odds ratios for survival rates in nitroimidazole cervix trials. Left hand columns represent author (year of publication), number of patients, odds ratio (OR), lower and upper 95% confidence limits (95% CI). Odds ratios of less than one suggest improvement of overall survival rates with the addition of a nitroimidazole.

Discussions with respect to toxicity occurred in all trials. However, some trials reported on all toxicity due to the radiation both with and without nitroimidazole, while others reported only on neurotoxicity, the dose-limiting toxicity related to nitroimidazoles. Two trials reported on both [14, 15]. In the MRC trial of pimonidazole [11], toxicity was reported by group and by manifestation. Unfortunately, it was not possible to delineate individual patients and many patients would likely be counted twice if a summated proportion were taken of patients with moderate to severe toxicity. This trial was not included in pooling of data for toxicity. Radiation-related toxicity was largely injury to the bladder, vagina or rectum. Only severe, life-threatening or fatal complications were included for pooling. Three trials reported radiation-related toxicity in enough detail to justify inclusion for pooling. As some cells contained values of zero, 0.5 was added to all cells. The resulting odds ratio for significant toxicity in the treatment group was 1.65 (95% CIs: 1.10, 2.48; p=0.016) with no evidence for heterogeneity within trials (p=0.37). Three trials reported specifically on neuropathy. The pooled odds ratio for the development of moderate, severe or life-threatening toxicity was 3.21 (95% CIs: 1.36, 7.55). Again, this reached statistical significance (p=0.008) with no evidence of heterogeneity (p=0.66).

Exploration of heterogeneity
No subgroup analysis was possible on publication status as all trials were reported as full journal articles. Individual patient data were not available, to study the effect of stage directly. However, studies were ranked by the severity of case mix. Two trials included only stage III and/or IV patients [10, 12, 13]. These two trials were compared with the remaining three trials, which included stage II patients. Odds ratios and 95% CIs were 0.92 (0.56, 1.51) and 1.29 (0.92, 1.80), respectively.

Three different nitroimidazole compounds were used within the five trials. A comparison was made between those trials using misonidazole [10, 12, 13, 15], ornidazole [14] and pimonidazole [11]. The pooled odds ratio (95% CIs) for the misonidazole trials was 1.05 (0.76, 1.43). Results for the ornidazole trial were 0.79 (0.48, 1.31) and for the pimonidazole trial, 1.85 (1.13, 3.04). This suggested an inferior effect of pimonidazole, however, this was not statistically significant (p=0.20).

Trials were ranked by methodological rigour. The two lower scoring trials were pooled, as were the three higher scoring trials (Table 1). Higher scoring trials achieved a pooled odds ratio (95% CIs) of 1.18 (0.56, 2.47). Lower scoring trials achieved similar results: 1.09 (0.75, 1.57).

As suggested by overlapping confidence intervals across subgroups and non-significant ² for heterogeneity, none of the factors chosen a priori explained variability between studies. It is noted that for all main analyses, fixed and random effects models gave pooled estimates of treatment effects that were either identical or very similar, suggesting good agreement between trials.

	Discussion

Top Abstract Introduction Methods and materials Results Discussion Conclusion References

 
Pre-clinical examination of nitroimidazoles in increasing radiosensitivity of hypoxic cells brought great promise of clinical benefit. Results from randomized controlled trials in several disease sites, including cervix, have been disappointing. The above overview was performed to allow for increased power and precision of an estimated treatment effect. The rate of local tumour control was considered the primary outcome of interest because it is a clinically important outcome. In addition, it is perhaps the best measure of the ability of the compound to increase radiosensitivity as overall survival is affected by many other outcomes such as metastases, treatment toxicity and co-morbidities. The pooled random effects odds ratio for local control was 1.14 (95% CIs: 0.78, 1.66) suggesting no benefit to the addition of nitroimidazole to radiation. While the use of nitroimidazoles has been established with respect to biological principle, their clinical use in cervix cancer is not supported by this meta-analysis. Two possible reasons may account for this. First, these compounds demonstrated a significant degree of toxicity. There is an increase of radiation toxicity, as might be expected with a radiosensitizer. In addition, these compounds have radiation-independent toxicity, largely neurotoxicity. Both of these occurred to a significant degree (p<0.01) in the pooled analysis. Such dose-limiting toxicities may attenuate the usefulness of nitroimidazoles in the clinic, limiting the safe dose to below the dose required for sensitization. Second, the tumours of nitroimidazole trials have been specifically chosen for study because of evidence of hypoxia. This indicates a poor blood supply and as such, tumours may experience poor drug delivery, not allowing for active drug to reach the malignant cells. Poor blood supply to the tumours but not to surrounding normal tissue may also explain why a biological effect was seen with respect to toxicity but not disease control. Either of these two possibilities may explain, in part, the lack of clinical effect.

With respect to overall survival, a non-significant trend against the use of nitroimidazoles was found. This is not an unexpected finding in light of the lack of local control benefit. An increased rate of radiation-induced toxicity such as recto-vaginal fistulae in addition to nitroimidazole-induced neurotoxicity may account for increased death rates in a population that has received no additional benefit.

While none of the five included studies showed benefit from the use of nitroimidazoles, there is evidence from a large randomized trial supporting the use of nimorazole in head and neck cancer [16]. One potential explanation for this apparent difference is that nimorazole may be relatively less toxic than other nitroimidazoles, allowing for higher doses [17]. As well, it may have better properties of tumour penetration than misonidazole [2], further enhancing its potential therapeutic ratio. This compound has not been studied in cervical cancer.

Reporting of outcomes for the five identified trials varied considerably. In some cases (e.g. toxicity), this prevented pooling from all trials. In other cases, such as survival, certain assumptions were made to allow for pooling because numbers of patients at risk were not always identified. For instance, where possible, survival curves were read from the time of minimal follow up, rather than the reported time point in order to ensure that all enrolled patients were at risk for an event. In addition, reported outcome time points varied across trials. To circumvent this, the assumption was made that relative rates of events between groups did not vary across time. The addition of such assumptions requires caution with respect to the interpretation of the results. While confidence intervals have been reported, these are purely statistical and based upon the above assumptions. Clinical intervals of confidence are likely wider. To overcome these difficulties in future analyses of trials, it is suggested that oncology journals adopt a series of standards for reporting, allowing for facilitation of data extraction and statistical pooling.

No significant heterogeneity between groups was found with respect to local control, by examination of the following factors: severity of disease, type of nitroimidazole used and methodological rigour. Possible alternative explanations include low power because of the inclusion of only five trials.

	Conclusion

Top Abstract Introduction Methods and materials Results Discussion Conclusion References

 
No significant benefit was found from the addition of a systemic nitroimidazole compound to a radical course of radiotherapy for carcinoma of the cervix when data from five trials were pooled. Rather, there was weak evidence for a detrimental effect on overall survival. Recently, a meta-analysis was performed supporting the use of concomitant chemotherapy with radiation for treatment of carcinoma of the cervix [6]. In light of this finding and the presented meta-analysis, further clinical study into the use of nitroimidazoles with radiotherapy for the treatment of cervical cancer must be approached with caution and even then, only if compelling evidence to support their further use is provided.

	Acknowledgments

 
Drs J Overgaard and S Dische are thanked for their communication during the identification of trials. Drs D Cook, G Guyatt and S Daya are acknowledged for their assistance during all phases of this study. J Julian is thanked for assistance with the figures.

	Footnotes

 
This work was performed, in part, through the financial assistance of the Juravinski Regional Cancer Centre Foundation. The authors performed all data collection and manuscript preparation.

Received for publication June 14, 2004. Revision received September 27, 2004. Accepted for publication October 19, 2004.

	References

Top Abstract Introduction Methods and materials Results Discussion Conclusion References

 

1. Adams GE, Dische S, Fowler JF, Thomlinson RH. Hypoxic cell sensitisers in radiotherapy. Lancet 1976;1:186–8.[Medline]
2. Overgaard J. Clinical evaluation of nitroimidazoles as modifiers of hypoxia in solid tumors. Oncol Res 1994;6:509–18.[Medline]
3. Sterne JA, Gavaghan D, Egger M. Publication and related bias in meta-analysis: power of statistical tests and prevalence in the literature. J Clin Epidemiol 2000;53:1119–29.[CrossRef][Medline]
4. Moher D, Pham B, Jones A, Cook DJ, Jadad AR, Moher M, et al. Does quality of reports of randomized trials affect estimates of intervention efficacy reported in meta-analysis? Lancet 1998;352:609–13.[CrossRef][Medline]
5. Detsky AS, Naylor CD, O'Rourke K, McGeer AJ, L'Abbe KA. Incorporating variations in the quality of individual randomized trials into meta-analysis. J Clin Epidemiol 1992;45:255–65.[CrossRef][Medline]
6. Green J, Kirwan J, Tierney J, Symonds P, Fresco L, Williams C, et al. Concomitant chemotherapy and radiation therapy for cancer of the uterine cervix (Cochrane Review). Cochrane Database Syst Rev 2001;4:CD002225.
7. Fleiss JL. The statistical basis of meta-analysis. Stat Meth Med Res 1993;2:121–45.[Medline]
8. Thornton A, Lee P. Publication bias in meta-analysis: its causes and consequences. J Clin Epidemiol 2000;53:207–16.[CrossRef][Medline]
9. Colton T. Statistics in medicine. Boston: Little, Brown and Co, 1974.
10. Dische S, Adams GE, Ash DV, Baker J, Barley V, Bleehen NM, et al. The Medical Research Council trial of misonidazole in carcinoma of the uterine cervix. (A report from the MRC Working Party on Misonidazole for Cancer of the Cervix.) Br J Radiol 1984;57:491–9.[Abstract/Free Full Text]
11. Dische S, Chassagne D, Hope-Stone HF, Dawes P, Roberts J, Yosef H, et al. A trial of Ro 03-8799 (pimonidazole) in carcinoma of the uterine cervix: an interim report from the Medical Research Council Working Party on advanced carcinoma of the cervix. Radiother Oncol 1993;26:93–103.[CrossRef][Medline]
12. Leibel S, Bauer M, Wasserman T, Marcial V, Rotman M, Hornback N, et al. Radiotherapy with or without misonidazole for patients withstage IIIb or stage IVa squamous cell carcinoma of the uterine cervix: preliminary report of a Radiation Therapy Oncology Group randomized trial. Int J Radiat Oncol Biol Phys 1987;13:541–9.[Medline]
13. Grigsby PW, Winter K, Wasserman TH, Marcial V, Rotman M, Cooper J, et al. Irradiation with or without misonidazole for patients with stages IIIb and IVa carcinoma of the cervix: Final results of RTOG 80-05. Int J Radiat Oncol Biol Phys 1999;44:513–7.[CrossRef][Medline]
14. Okkan S, Atkovar G, Sahinler I, Turkan S, Uzel R. A randomised study of ornidazole as a radiosensitiser in carcinoma of the cervix: long-term results. Br J Cancer 1996;74 (Suppl. 27):S282–6.
15. Overgaard J, Bentzen SM, Kolstad P, Kjoerstad K, Davy M, Bertelsen K, et al. Misonidazole combined with radiotherapy in the treatment of carcinoma of the uterine cervix. Int J Radiat Oncol Biol Phys 1989;16:1069–72.[Medline]
16. Overgaard J, Hansen HS, Overgaard M, Bastholt L, Berthalsen A, Specht L, et al. A randomized double-blind phase III study of nimorazole as a hypoxic radiosensitizer of primary radiotherapy in supraglottic larynx and pharynx carcinoma. Results of the Danish Head and Neck Cancer Study (DAHANCA) Protocol 5-85. Radiother Oncol 1998;46:135–46.[CrossRef][Medline]
17. Timothy AR, Overgaard J, Overgaard M. A phase I clinical study of Nimorazole as a hypoxic radiosensitizer. Int J Radiat Oncol Biol Phys 1984;10:1765–8.[Medline]

This article has been cited by other articles:

				J. Overgaard Hypoxic Radiosensitization: Adored and Ignored J. Clin. Oncol., September 10, 2007; 25(26): 4066 - 4074. [Abstract] [Full Text] [PDF]

				BJR Review of the Year - 2005. Br. J. Radiol., March 1, 2006; 79(939): 183 - 187. [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Dayes, I S Articles by Abuzallouf, S Search for Related Content PubMed PubMed Citation Articles by Dayes, I S Articles by Abuzallouf, S