Treatment of hepatocellular carcinoma: a systematic review of randomized controlled trials

Annals of Oncology 8:117-136,1997. © 1997 Kluwer Academic Publishers. Printed in the Netherlands.

Review Treatment of hepatocellular carcinoma: A systematic review of randomized controlled trials R. G. Simonetti,1 A. Liberati,2 C. Angiolini2 & L. Pagliaro3 l

Divisione di Medicina, Ospedale V. Cervello, Palermo; 2Laboratorio di Epidemiologia Clinica, Istituto M Negri Milan; * Istitulo di Medicina Generate e Pneumologia, Universita di Palermo, Italy

Summary

Introduction Hepatocellular carcinoma (HCQ is one of the most common cancers in males, with a world-wide incidence between 250,000 and 1,000,000 new cases per year, and a male-female ratio of about 4 or 5:1 [1-3]. The geographic distribution is highly uneven: hepatocellular carcinoma is less frequent in western countries [4] where it is also more clearly associated with advanced, clinically overt cirrhosis [5,6]. The tumor is more frequent in some African and Asian regions; in Africa it often occurs in young people and is associated with early, clinically inapparent cirrhosis [6]. Overall, HCC is associated with cirrhosis in about 80% of cases [6, 7]. Surgery is considered the treatment of choice for HCC, but only a small proportion of patients are candidate for radical resection at the time of diagnosis. For most non-resectable patients the need of an effective non-surgical treatment is obvious. Results of clinical trials so far carried out are inconclusive [8]. To our knowledge, however, no systematic evaluation of the methodologic quality and results of these trials has

Conclusions: This systematic review of RCTs on HCC, mostly in non resectable patients, indicate that the non-surgical current treatments are ineffective or minimally and uncertainly effective. The three treatment modalities minimally and uncertainly effective (i.e., embolization, tamoxifen and IFN) can deserve further assessment by larger and methodologically more sound randomized trials. Key words: hepatocellular carcinoma, meta-analysis, randomized controlled trials, treatment

been carried out. Therefore we undertook a systematic overview of available RCTs in HCC. We also discuss the implications of this review for the planning and design of future trials.

Methods Selection of randomized trials We restricted our attention to phase III RCTs. Studies were identified by searching through MEDLINE for specific key words (i.e., randomized controlled trial, clinical trials, random allocation, hepatocellular carcinoma, hepatoma) and by perusing the reference lists of original and review articles on the treatment of primary liver cancer. Trials were considered ineligible if they were preliminary or brief reports, if they included patients with metastatic cancer to the liver. Initially, 42 English-language papers were retrieved; 5 were then discarded for the following reasons: one described a trial published previously and already included in the search [% one because it reported the results from a large trial on Chinese HBsAg carriers in which a subset of patients with HCC was not separately analyzed [10]; two trials [11, 12] because they only reported preliminary results. The last [13] because it evaluated

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

Background: Hepatocellular carcinoma ( H C Q is a leading cause of cancer-related death. Many treatments have been proposed but considerable uncertainty still remains about their effectiveness. In this review we evaluated the quality, clinical coherence, consistency and results of Randomized Controlled Trials (RCT) of non-surgical treatments for HCC. Methods: Thirty-seven RCTs examining the effect of different treatments were retrieved using MEDLINE (November 1978 to December 1995) and a review of reference lists. Selected aspects of the quality of design, conduct and reporting were examined. The odds ratio for the probability of surviving up to one year was calculated according to the Mantel-HaenszelPeto method and displayed using l'Abbe plots. Results: The 37 RCTs overall included 2803 patients (median 56, range 20-289). Patients prognosis varied widely across studies which also failed to report on important information about their characteristics. Only 10 RCTs had an untreated control group; the remaining 27 compared different regimens of intravenous or intraarterial chemotherapy with or without embolization of hepatic artery, hormono- and immunotherapy

regimens. Some evidence of a moderate benefit emerged only from RCTs using tamoxifen and transcatheter arterial embolization vs. no treatment in unresectable patients: pooled odds ratio for 1-year survival were, respectively, 2.0 (95% confidence intervals (CI) 1.1-3.6) and 2.0 (95% CI 1.1-3.6). At 2 years, however, pooled odds ratio were no longer statistically significant for tamoxifen 1.2 (95% CI 0.6-2.6) but was significant for embolization 2.3 (95% CI 1.2-4.6). No evidence of efficacy was detected for embolization as adjuvant therapy in resected or transplantated patients nor for chemotherapy added to intraarterial embolization.

118 Table 1. General characteristics of the studies reviewed. 37 2803 1974-1994 1978-1995 11 18 8 2 17 11 1 6 9 9 7 10 2 10 23 2 1 2

maintenance therapy (Biological Response Modifiers) after treatment with percutaneous ethanol injection, transcatheter arterial embolization or arterial infusion of antitumor agents. Eventually, 37 RCTs published between November 1978 and December 1995 reporting results on 2803 patients were eligible for this review.

Review of the trials We analyzed selected aspects of the methodology and completeness of reporting of the studies. With regard to the methodologic quality we looked at the following aspects of study design and conduct: a) method of randomization; b) criteria for response assessment; c) patients dropped out or withdrawn from the study before its termination; d) methods used to handle drop-outs and withdrawals in the analysis; e) evidence of an a priori estimate of sample size. Completeness of reporting was assessed looking at: a) information on patient characteristics; b) description of therapeutic regimen; c) timing of events; d) reporting of side-effects. Operational definitions of the criteria adopted to analyze each item are reported in Appendix 1. To look at chronologic development of hypotheses tested and to identify whether promising areas for future research exist, studies were classified into five groups according to the prevailing therapeutic combinations tested a) chemotherapy; b) hormonotherapy, c) embolization; d) immunotherapy; e) other. If combination therapy was used and the comparison was 'treatment A' + else versus the same else, trials have been considered testing the effectiveness of treatment 'A'. For example the trial by Melia et al. [14] in which tamoxifen plus doxonibicin was compared to doxorubicin alone, was classified in the group 'hormonotherapy'.

Statistical analysis One-year survival, computed from life-table or sometimes from the curves plotted in the papers was our primary outcome measure. It was chosen because survival at this time was reduced to 50% or less in most studies. The odds ratios with 95% CI of one-year survival rate in treated patients compared with controls in each trial (in every arm if there were

Results General characteristics of the studies Characteristics of the 37 RCTs included in this review are listed in Table 1. Twenty-six trials were multicentric. Hepatologists, Surgeons, Oncologists and Internists were the specialists more frequently involved. The median number of subjects enrolled was 56 (range = 20 to 289). Most RCTs compared active treatments one another without an untreated arm. The large variability in the baseline prognosis of patients enrolled in different studies is apparent looking at the outcome of the control arms or RCTs where an untreated group was maintained (Figure 1). Design and execution (Table 2) In 7 RCTs treatment allocation was carried out via a fully blinded randomization, in 10 studies sealed envelopes were used. In 1 trial patients were assigned according to even or odd number or by data of birth; while in 19 RCTs the method of randomization was not specified. Across the trials, partial or complete objective response was assessed using different, often poorly reliable, criteria such as hepatomegaly, tumor size, alphafetoprotein, and presence of metastases. Definition of response was highly variable even when the same criterion was measured across trials. Decrease in tumor mass was inferred from the disappearance or size reduction of palpable liver or directly measured by ultrasound or computer tomography with or without previous lipiodol angiography, or at autopsy. Similarly, decrease in AFP level was reported when the changes were 50%, or > 25%-3O%. Because of these inconsistencies we did not use 'response' as an outcome measure. Eleven RCTs reported that no patient were lost to follow-up, 11 excluded patients from the analysis after randomization; 12 papers did not give information about withdrawal analysis. Only 10 RCTs had an a priori calculation of the

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

Number of patients Number of patients Period of Conducting Reporting Participation center for each study Single < 5 > 5 Country of origin Africa Asia Europe USA Africa-USA Principal investigator specialty Hepato-gastroenterology Internal medicine Oncology Surgery Radiology Type of journal Hepato-gastroenterology Oncology Surgery Medicine Radiology

multiple comparisons) were computed by the Peto-Mantel-Haentzel method, as reported by Collins et al. [15]. Odds ratios greater than unity representing a beneficial effect of the experimental treatment are shown in plots in which horizontal bars representing 95% CI that do not cross the vertical equivalence line correspond to a statistically significant treatment effect In the plots the trials are displayed according to the year of publication. Pooled odds ratios (P-OR) were computed when there was a similarity between the therapeutic regimens across the trials within each set. Intertrial heterogeneity in treatment effect was evaluated using both a chi-square test and a visual display in a l'Abbe plot [16]. In the l'Abbe plot the co-ordinates of each dot correspond to the outcome rates in each treatment group of each RCT. The diagonal correspond to the line of identity or equivalence between treatments. Dots below or above the identity line indicate that the treatment represented on the horizontal axis is better or, respectively, worse than the treatment represented on the vertical axis. A sensitivity analysis was performed when appropriate. We reported also median survival figures (Tables 4-10), and odds ratios of the 2-year survival figures when available and contributory.

119 Table 2. Selected characteristics of design, execution and reporting. Randomization method

Response evaluation

Drop-outs, withdrawals and protocol deviations

Handling of drop-outs and withdrawals

A priori estimate of samplesize

1. Falkson[17]

1978

Not reported

7 not eligible 5 not evaluable

Excluded

No

2. Melia[18]

1983

Not reported

None

-

No

3. Choi [19]

1984

Not reported

Hepatomegaly 1 Tumor mass 1 Lung metastases 2 Hepatomegaly 1 Tumor mass 2 AFP levels Hepatomegaly 1

Excluded

No

4. Falkson[20]

1984

Not reported

Hepatomegaly 1 Tumor mass 1 Lung metastases 2

Excluded

No

5. Falkson[21]

1984

Not reported

Not reported

Excluded

No

6. Bezwoda[22] 7. Falkson[23]

1987 1987

Sealed envelopes Not reported

Excluded

No Yes

8. Lai [24]

1988

Sealed envelopes

-

No

9. Falkson[25]

1990

1992

Ineligible patients were included for toxicity assessment -

Yes

10. Kajanti[26]

Central office (according to strata) Sealed envelopes

Hepatomegaly 1-2 Hepatomegaly 1 Tumor mass 1 Tumor mass 2 Lung metastases 2 AFP levels Tumor mass 1-2 synthoms

1 not eligible 1 refusing treatment 19 withdrawals 21 not eligible 1 protocol deviation 1 drop-out 1 withdrawals 14 not eligible None 4 withdrawals 13 not eligible None

11. Melia[14]

1987

Sealed envelopes

2 withdrawals 4 drop-out

Not evaluable

No

12. Farinati[27] 13. Uchino[28]

1990 1993

Not reported Central office

Not reported 4 withdrawals

Not evaluable Excluded

No No

14. Elba [29] 15. Martinez-Cerezo [30] 16. Manesis[31]

1994 1994 1995

Not reported Sealed envelopes Sealed envelopes

Not evaluable Excluded

No Yes Yes

17. Castells[32]

1995

Tumor size 2

Excluded

Yes

18. Lin [33]

1988

Computer-generated after stratification according to Okuda's stage Not reported

Not reported None 1 withdrawal 4 drop-outs 2 withdrawals (side effects)

Not evaluable

No

19. Pelletier [34]

1990

Sealed envelopes

-

Yes

20. Yoshikawa [35] 21. Madden [36]

1994 1993

Not reported Sealed envelopes

Excluded Intention to treat

No Yes

22. Groupe Francophone HCC [37] 23. Kawai[38]

1995

Central office

Intention to treat

Yes

1992

Central office

Intention to treat

No

24. Chang [39] 25. Ikeda[40]

1994 1992

Not reported Not reported

Not reported Not reported

Not evaluable Not reported

No No

26. Watanabe [41] 27. Ikeda[42] 28. Izumi[43]

1994 1995 1994

Sealed envelopes Not reported Not reported

Tumor mass 2 AFP levels Tumor size 2 Tumor size 2 AFP levels Tumor size 2 AFP levels Tumor size 2 AFP levels Lipiodol accumulation (CTscan) Lipiodol retention Tumor size 2 AFP levels Tumor size 2 % tumor necrosis AFP Examination residual liver 2

7 withdrawals 2 drop-out 1 protocol deviation None

9 withdrawals 1 withdrawal 1 refusing treatment 1 withdrawal (poor liver function)

Not evaluable Not evaluable Excluded

No No No

Tumor mass AFP levels enzymes Hepatomegaly 1 Tumor size 2 AFP levels Tumor size 2 AFP levels Not reported Tumor size 2 Tumor size 2

Tumor mass 2 AFP levels

2 withdrawals 17 ineligible None

2 withdrawals 7 withdrawals 3 drop-outs 2 protocol deviation 1 withdrawals 2 not eligible 23 protocol deviation 4 early deaths

No

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

Publication (year)

Authors [References]

120 Table 2. Continued. Publication (year)

Randomization method

Response evaluation

Drop-outs, withdrawals and protocol deviations

Handling of drop-outs and withdrawals

A priori estimate of samplesize

29. Wu [44]

1995

Not reported

Not reported

Not reported

No

30. Lygidakis [45]

1995

Not reported

Not reported

Not reported

No

31. Lai [46]

1989

Day of histological diagnosis

None

-

No

32. Lai [47]

1993

Sealed envelopes

None

-

m

33. Falkson[48]

1995

Central office (according to strate)

AFP examination residual liver 2 CEACA19-9, CA-50, AFP TCScan Hepatomegaly 1 Tumor mass (2 or at death) AFP or ferritin levels Hepatomegaly 1 Tumor mass (2 or at death) AFP and/or ferritin levels Not reported

4 withdrawals 4 ineligible

Yes

34. Kawata[49]

1995

Not reported

None

Excluded (4 ineligible mcluded for toxicity assessment) -

35. Lai [50]

1986

Not reported

None

-

36. Order [51] 37. Raoul[52]

1991 1994

Central office Not reported

None Not reported

Not reported

Recurrence of liver tumor 2 AFP Hepatomegaly 2 and/or at death Tumor mass 2 AFP Evolution portal vein thrombosis

No

No Yes

The number given on the left of the name of the first author of each trial is the same used in Tables 3-10 and in Figures 2-6. 1 = Clinical diagnosis; 2 := by laboratory investigations or imaging; AFP = alpha-fetoprotein; CT = computed tomography.

100 + 34, Lai • 8. Lai - 12. Fannati • 19. Pelletier 3 32. Lai ••" 15 Martinez -*14. Elba -=• 16. Manesis • 22. Groupe F. HCC -* 17. Castells Figure I Differences in survival of control groups used in trials testing the effectiveness of various experimental treatments.

number of patients needed to detect a clinically and statistically significant treatment effect: 2 of them however did not specify the expected rate of events on which the sample size was estimated. In another trial a sequential design was used.

Reporting (Table 3) Published reports were particularly unsatisfactory in the description of the characteristics of the patients: only 7 RCTs gave complete information about baseline patients

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

Authors [References]

121 Table 3. Selected aspects of the reporting of the studies reviewed. Regimen description

Timing of events

Discussion of side-effects

Note

1. Falkson[17]

No information

Full information

Complete information

- Unresectable

2. Melia[18]

Full information

Complete information

- Unsuitable for resection or transplantation - 16 Patients previously treated by selective hepatic arterial embolization

Full information

Survival curve without number of patients at risk

Complete information

- Unresectable

4. Falkson[20]

Age Gender Cirrhosis (yes/no) Bilirubin Performance status Previous treatment Age Gender No previous treatment Bilirubin Performance status No information

Survival curve with number of patients at risk at onset and at end Survival curve without number of patients at risk

Full information

Complete information

- Unresectable

5. Falkson[21]

No information

Full information

Survival curve with the number of patients at risk at onset and the total number of events Survival curve without number of patients at risk and with the total number of events

Complete information

6. Bezwoda[22]

Age Gender Albumin, bilirubin Performance status No previous treatment Age Gender Cirrhosis (yes/no) Metastases Syntoms

Only information on dosage

Survival curve without number of patients at risk

Full information

Survival rate

Only qualitative information + number of deaths drug correlated Complete information

- Excluding patients with advanced liver disease. 9 Patients where entered on the cross-over step of the protocol; including only patients with measurable tumor mass, in performance status 0-3 and adequate hematologic and renal function; stratified by geographic origin - Unselected - excluding patients with PSE,' poor performance status, UGI" bleeding

Age Gender No previous treatment Bilirubin Performance status Gender Syn thorns Performance status Jaundice (yes/no) Metastases

Full information

Survival curve without number of patients at risk

Complete information

Full information

Survival curve without number of patients at risk

Complete information

Full information

Survival curve without number of patients at risk

Not reported

Only information on dosage and scheduling Full information

Survival curve without number of patients at risk Survival curve without number of patients at risk

Not reported

- Unselected 6 patients previously treated by TAE

No side effects

- Unresectable

3. Choi [19]

7. Falkson[23]

8. Lai [24]

9. Falkson[25]

10. Kajanti[26]

11. Melia[14]

12. Farinati[27]

Age Gender No previous treatment Non-cirrhotic patients Liver function Clinical stage Performance status Age Gender Cirrhosis (yes/no) Age Gender Cirrhosis (yes/no) Clinical stage Tumor mass

• Also resectable patients? Including only patients with measurable tumor mass, in performance status 0-3 and adequate hematologic and renal function • Unresectable

• Including only patients with measurable tumor mass, in performance status 0-3 and adequate hematologic and renal function, without encephalopathy and recent bleeding

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

Patient characteristics

Authors [References]

122 Table 3. Continued. Patient characteristics

Regimen description

Timing of events

Discussion of side-effects

Note

13. Uchino[28]

Age Gender No previous treatment Child's classification Clinical stage Age Gender Child's classification Clinical stage Tumor mass Age Gender Chronic liver disease No previous treatment Child's classification Performance status Tumor size and number Age Gender Cirrhosis No previous treatment Child-Pugh classification Clinical stage Performance status Tumor size Age Gender Cirrhosis (yes/no) No previous treatment Child-Pugh classification Clinical stage Performance status Tumor stage (solitary/ multiple/massive) Age Gender Albumin bilimbin

Full information

Survival curve without number of patients at risk

Complete information

- Unresectable

Only dosage

Survival curve without number of patients at risk

Complete information

- Unresectable non eligible for alcohol injection or chemoembolization

Only dosage

Survival curve without number of patients at risk

Complete information

Full information

Survival curve without number of patients at risk

Only qualitative information

- Including patients with advanced cancer (tumor diameter > 5 cm multiple nodules, portal thrombosis); exluding patients in poor condition and with metastasis - Unresectable, including patients without bleeding and or PSE" and or m 4 °C grade ecog scale

Full information

Survival curve with number of patients at risk at every different time-points

Complete information

- Unsuitable for resection, OLT, alcoholization or transarterial embolization; excluding patients with uncontrolled decompensation, PSE,' UGI b bleeding or infection

No information

Life table with number of patients at risk at every different timepoints

Complete information

Full information

Survival curve without number of patients at risk

Only qualitative information

- Unresectable, excluding patients with hepatic decompensation, main portal vein thrombosis, extrahepatic metastases - Unresectable, including compensated and decompensated patients

Full information

Survival curve without number of patients at risk

Complete information

- Unresectable, including compensated and decompensated patients, good performance status, some with portal vein thrombosis

Only information on dosage and scheduling

Survival curve with number of patients at risk at every different time-points

Not reported

Only information on dosage and scheduling

Survival curve without number of patients at risk

Complete information

- Unresectable, including compensated and decompensated patients; excluding patients with metastasis and in poor condition - Unresectable, including compensated patients (Child's class A), 7 patients with segmental portal obstruction; exluding patients with metastasis

14. Elba [29]

15. Martinez-Cerezo [30]

16. Manesis[31]

17. Castells[32]

18. Lin [33]

19. Pelletier [34]

20. Yoshikawa [35]

21. Madden [36]

22. Groupe Francophone of HCC [37]

Age Gender Cirrhosis (yes/no) Albumin, bilimbin Prothrombin time Clinical stage Tumor size Age Gender Cirrhosis Child's classification Clinical stage Performance status Age Gender Clinical stage Performance status Age Gender Cirrhosis No previous treatment Child's classification Clinical stage Tumor size and number

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

Authors [References]

123 Table 3. Continued. Authors [References]

Patient characteristics

Regimen description

Timing of events

Discussion of side-effects

Note

23. Kawai[38]

Age Gender Cirrhosis (yes/no) No previous treatment Child's classification Clinical stage Performance status Tumor size and picture Age Gender No previous treatment Child's classification Tumor picture (single or multiple nodules, diffuse type) Age Gender

Full information

Survival curve without number of patients at risk

Complete information

Also resectable patients? Mostly patients are in Child's A or B class

Only information on dosage and scheduling

Survival curve without number of patients at risk

Complete information

- Unresectable

Incomplete information

Survival curve without number of patients at risk Survival curve without number of patients at risk

Complete information

Including various stages of HCC

Incomplete information

- Unresectable

24. Chang [39]

25. Ikeda[40]

26. Watanabe [41]

28. Izumi[43]

29. Wu [44]

30. Lygidakis [45]

31. Lai [46]

32. Lai [47]

33. Falkson[48]

34. Kawata[49]

Age Gender TNM stage Cirrhosis

Only dosage

Only information on dosage and scheduling

Survival curve without number of patients at risk

Complete information

- Unresectable? - Performance status 01 - Good liver function

Full information

Survival curve without number of patients at risk

Incomplete information

Postoperative adjuvant therapy

Full information

Survival curve with number of patients at risk at every different time-points - Survival rate - Survival free disease rate

Incomplete information

Preoperative adjuvant therapy for resectable large ( ^ 10 cm) tumor

Complete information

- Including patients without extrahepatic disease, tumormass < 50% of the liver surface, Okuda's stage 1-2 - Unresectable

Full information

Full information

Survival curve without number of patients at risk

Complete information

Full information

Survival curve without number of patients at risk

Not reported

- Unresectable, no metastases

Full information

Survival curve with number of patients at onset and total number of events

Complete information

Full information

Survival curve without number of patients at risk

Complete information

- Including patients with unresectable, residual, recurrent or metastatic measurable area of disease: performance status 0-3, adequate hematologic renal and liver function - Stratified by geographic origin Resection of the lesion, good or fair performance status, absence of metastases

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

27. Dceda[42]

Age Gender Clinical stage Child's classification Tumor size and number Performance status Age Gender Cirrhosis (yes/no) No previous treatment Decompensation Performance status Tumor size and number Age Gender Cirrhosis (yes/no) Tumor size and number Age Gender Cirrhosis (yes/no) Child's classification Age Gender Okuda's stage TNM classification Age Gender No previous treatment Bilirubin, ascites Performance status Age Cirrhosis present or absent Bilirubin, albumin -

124 Table 3 Continued. Authors [References]

Patient characteristics

Regimen description

Tuning of events

Discussion of side-effects

Note

35. Lai [50]

Age Gender No previous treatment Performance status Age Bihrubin

Full information

Survival curve without number of patients at risk

Complete information

- Unresectable

Full information

Survival curve without number of patients at

Complete information

- Unresectable, no 'gross' ascites

Age Cirrhosis (yes/no) Child's classification Clinical stage Tumor number

Full information

Survival curve without number of patients at

Complete information

Portal vein thrombosis without metastases, stage 1 or 2 Okuda

36. Order [51]

risk 37. Raoul[52]

risk

" PSE = portal-systemic encephalopathy. b UGI bleeding = upper gastrointestinal bleeding.

Analysis of clinical results Chemotherapy Ten RCTs evaluated chemotherapeutic regimens (Table 4), 7 including doxorubicin (ADM) and 2 studies using doxorubicin-analogue. Doxorubicin, however, was compared with a non-active treatment in only 1 trial [24]. In 6 RCTs it was tested against one or more regimens not containing doxorubicin [17-19, 21-26] or versus multiple regimens, one of which included doxorubicin [20]. According to these characteristics, we considered ADM or analogue as the 'experimental' treatment. At 1 year doxorubicin was significantly more effective than 5-fluorouracil (5FU) [17]. It was slightly more effective (but without reaching statistical significance) compared to no active treatment [24], added to VM 26 and 5-FU versus a combination [22], versus neocarzinostatin and versus m-AMSA [21]. Conflicting results were obtained when doxorubicin was compared with 5-FU plus methil-CCNU and with 5-FU plus streptozocin in the two studies performed by Falkson [17, 20]. No differences were found between 4'deoxydoxorubicin and acivin [25]. In the only trial in which no doxorubicin was used [23] no differences in survival emerged between mitoxantrone and cisplatin. At 2 years the differences between treatments in the trials were further reduced. Very few patients were actually observed even when the curves report data until 2 years. Finally, in a study evaluating intraarterial versus intravenous 5-FU + 4 epidoxorubicin [26], no difference was found.

The heterogeneity of trials design and of the clinical hypotheses (with doxorubicin tested against a wide variety of drugs and regimen) prevented from a statistical and graphical display pooling of the results. Side effects were frequently seen with each of the treatments. In some cases drug-related deaths were reported - for cardiotoxicity after ADM administration and infections in neuthropenic patients after 5-FU alone or with streptozocin and in other combination regimens. Adverse reactions reducing the quality of life such as nausea, vomiting and alopecia (present in near 100% of cases treated by ADM) were often reported. The incidence of severe adverse reaction is reported in Table 4. Hormonotherapy Tamoxifen has been the most frequently studied drug (Table 5). In five studies [27, 29-32] it was compared to a non-active treatment (in one study tamoxifen was associated to the luteinizing hormone-releasing hormone-analogue triptorelin). The pooled odds ratio of survival at 1 year for all 7 RCTs was statistically significant (2.2, 95% CI 1.4-3.5) without statistical (P = 0.5) or visual heterogeneity (Figure 2). No statistically significant differences emerged at 2 years (pooled OR 1.5, 95% CI 0.8-2.7). Among the individual RCTs assessing tamoxifen alone versus non-active treatment, a borderline statistically significant therapeutic benefit was found only in 1 out of four studies (at 1 year) (Figure 3). The pooled odds ratio in this sub-group of studies showed borderline positive result at 1 year (2.0, 95% CI 1.1-3.6) (heterogeneity chisquare, P = 0.3, visual display in Figure 3). No differences were found at 2 years (pooled OR 1.2, 95% CI 0.6-2.0). Side effects were rare and self-limiting. Embolization We identified 3 groups of studies. In the first and third group, transcatheter arterial embolization (TAE) is assessed as palliative treatment in unresectable patients and as adjuvant therapy in resectable patients, respectively. In the second group the role of chemotherapeutic agents added to TAE (CHEMO-TAE) was analyzed.

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

characteristics such as age, gender, presence of cirrhosis, liver dysfunction, tumor size and previous treatments. Only 5/7 RCTs also reported information about performance status in the different arms of the trials. Nineteen out of 37 papers did not report on the presence/absence of cirrhosis in trial patients. Occurrence of side effects was discussed in 32/37 papers. Precise and quantitative information on the type of side effects was however present in 27/32 RCTs, with the remaining 5 studies only addressing the issue in general terms.

125

20 Vosfttatt 21 U a o w " 22 G'OiW F HCC

r 26 Wraraot 27 60 years old)

a) Mitoxantrone 14 mg/mq i.v. (?) b) Cisplatin 75 mg/mq i.v. (?)

14 14

54(16-80)

a) Doxorubicin 60 mg/mq, i.v. (60) b) No antitumor therapy (46)

10.6 7.5

106

-

9. Falkson[25]

75

Not evaluable

a) 4T>eoxydoxorubicin 30 mg/mq i.v. (30) b) Acivin 20 mg/mq i.v. (26)

21.4 11.6

10. Kajanti[26]

20

53(21-85)

a) 4'Epidoxorubicin + 5-FU 500 mg/mq intra-arterial (10) b) 4'Epidoxorubicin + 5-FU 500 mg/mq i.v. (10)

15.2 13.8

* Only patients without prior chemotherapy. b Unknown in the whole group. c Patients were stratified by geographic origin: North-American/European patients were randomized to neocarzinostatin or m-AMSA, SouthAfrican patients were randomized to neocarzinostatin or doxorubicin. Nine patients entered the cross-over step. d As indicated by authors, and/or inducing treatment withdrawal or modification of scheduling according to intention to treat. e Eight-month survival. - Not reported.

study [34] about 50% of patients were ascitic; in the Madden's study [36] the large majority were in Okuda prognostic stage 2 and 3; in the Yoshikawa's study [35] about 30% of the patients were in Child's C class. The other 2 studies included mostly compensated patients. Furthermore, a large variety of embolizing agents (with or without lipiodol) and of scheduling were used. For the whole group of studies (including RCTs with treated or not treated patients in the control arm) the pooled odds ratio was 2.0 (95% CI 1.2-3.3). Heterogeneity chi-square P = 0.27. In the set of RCTs where embolization was compared to no treatment or to intravenous 5-FU (assuming it not efficacious, see chemotherapy) the pooled

OR of 1-year survival rates showed a border-line significant therapeutic benefit (OR = 2.0, 95% CI 1.1-3.6). Two of the three RCTs obtained favourable results (significant in one), whereas the third was marginally on the unfavourable side of the equivalence line (Figure 5). The results are not modified by also summing up the results of the Madden study, in which survival is computed at 6 months. In 3 studies, 2-year survival was reported, with a slightly significant higher survival in the treated patients (pooled OR = 2.3, CI 95% 1.2^.6). However, given the heterogeneity of the design of the trials, of the characteristics of the included patients, of the used treatments and of the modalities of application across the trials, the

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

8. Lai [24]

15 12 13

127

1-Year survival

Severe adverse reaction d

2-Year survival

Treatment related death

Treated

Controls

Odds ratio (95% CI)

(a) 34* (a) 34* (a) 34*

(b) 5* (c) 25* (d)16*

8.06 (2.3-28.2) 1.50(0.5-^.5) 2.7(0.9-8.1)

(3)3?

(b)ll«

3.1 (0.7-13.6)

a) Cardiotoxicity 5 b) Anemia 14, 1$ WBC 14

(a) 5

(b) 2.5

0 9(0.1-15.7)

a) (| WBC 25, Cardiotoxicity 2 b) V WBC 16

(a) 9 (a) 9 (a) 9

(b)24 (c)18 (d)24

0.4(0.1-1.1) 0.5(0.1-1.5) 0.3(0.1-1.1)

Anemia a) 7 b) 9 c) 2 d)

(a) 15 (a) 15 (b) 8

(b) 8 (c)10 (c)10

2.2(0.4-12)* 1.6(0.3-7.6) 0.7(0.1-4.3)

a)b) I) WBC 11 PLT, neurological toxicity: 39 c) 1] WBC l± PLT: 58

(a) 21

(b) 9

2.5 (0.5-12.5)

Anemia a) ? b)?

Treated

Controls

Odds ratio (95% CI)

(a) 3

(b) 2

Anemia 22 0 0 0

U WBC 9 12 8 5

V WBC 21 2 7 5

V WBC ? ?

I) PLT 0 8 0 0

V WBC a) 34

(a) 23

(b)23

1.0(0.3-3.5)

(a) 10

(b) 10

1 (0.06-17.2)

cumulative evaluation of the efficacy of embolization must be viewed with great caution. The second group of these studies (Table 7, Figure 4) included 5 RCTs overall. In two studies chemotherapic agents (respectively ADM and cisplatin) were added to lipiodol transcatheter arterial embolization and compared to TAE. Conflicting results were found, with a non significant trend toward a better response at 1 and 2 years by using ADM + TAE in one study and a worse prognosis in the group treated by cisplatin + TAE in the other study, compared to patients treated only by TAE (1-year survival, pooled OR 1.2 95% CI 0.7-1.9). In the other 3 RCTs, several drugs were compared when added to TAE.

Cardiotoxicity 13

4 6 Cardiotoxicity 5

(( PLT 2 29 8 16

V PLT ? ?

a) li WBC, sepsis b) Neurological effect and cardiotoxicity Sepsis + WBC 1.5(0.1-15.2)

Infection

Infection 15

a) (J WBC V PLT: 8 Hematological and neurological: 37 b) GI, hematological and neurological: 48

Infection 8 13 Cardiotoxicity 1/?

Cardiotoxicity 10 Sepsis 3

No differences were found (pooled OR = 1.0, 95% CI 0.6-1.7). Finally, in the third group (Table 8, Figure 4) only for two studies [43-44] in which TAE was evaluated as adjuvant therapy used pre or post-operatively differences in survival could be evaluated. There was a trend towards a worse prognosis in treated patients (pooled OR = 0.4, CI 95% 0.2-1.0, heterogeneity chi-square P = 0.1). In the third study [45], fewer recurrences and a significantly higher recurrence-free survival are reported in the patients treated by TAE. In studies of TAE and Chemo-TAE where side-effects were thourougly discussed they were quite frequent. The

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

Not evaluable

a) b) c) d)

128 Table 5. Patients characteristics, treatments and outcome rates of each trial. Hormonotherapy studies. Authors [References]

Number of patients

Mean age (range)

Treatment in each arm (number of patients for arm)

Median survival (weeks)

11. Melia[14]

59

52 (20-73)

a) Doxorubicin 60 mg/mq + tamoxifen 20 mg/day (29) b) Doxorubicin 60 mg/mq (30)

12. Farinati[27]

38

64

a) Tamoxifen 30 mg/day (19) b) No treatment (19)

13. Uchino[28]

30

59 (31-75)

a) Cisplatin 60 mg/mq + doxorubicin 13 mg/mq intra-arterially + oral 5-FU 150 mg/mq/day + tamoxifen 25 mg/mq/day and medroxy-progesterone 40 mg orally (15) b) Cisplatin 60 mg/mq + doxrubicin 13 mg/mq intra-arterially + oral 5-FU 150 mg/mq/day (15)

14. Elba [29]

22

Not evaluable

a) Tamoxifen 60 mg/day (11) b) No antitumor therapy (11)

74 52

15. Martinez-Cerezo [30]

36

66

a) Tamoxifen 20 mg/day (20) b) No antitumor therapy (16)

37 25

16. Manesis[31]

85

62

a) Tamoxifen 30 mg/day + triptorelin 3.75 mg l.m./monthly (33) b) Flutamide 750 mg/day + triptorelin 3.75 mg i.mVmonthly (23) c) Placebo (29)

40* 16* 18*

17. Castells[32]

120

65

a) Tamoxifen 20 mg/day (58) b) Placebo (62)

6 8 36 8

Table 6. Patients characteristics, treatments and outcome rates of each trial. Embolization studies. Authors [References]

Number of patients

Mean age (range)

Treatment in each arm (number of patients for arm)

18. Lin [33]

63

49 (30-70)

a) Multiple TAE (ivalon + gelfoam) sessions + monthly i.v. 5-FU 1 g/mq (21) b) A single TAE (ivalon + gelfoam) session + monthly i.v. 5-FU 1 g/mq (21) c) Monthly i.v. 5-FU 1 g/mq (21)

19. Pelletier [34]

42

64 (41-82)

a) Chemo-TAE (gelfoam + doxorubicin) 50 mg/mq (21) b) No treatment (21)

20. Yoshikawa [35]

38

59 (38-75)

a) Chemo-L-TAE (4'epidoxorubicin 70 mg + lipiodol 2-3 ml), every 4 weeks (19) b) Intrahepatic arterial infusion of 4'epidoxorubicin (70 mg) every 4 weeks (17)

21. Madden [36]

50

(18-70)

a) Chemo-L-TAE (5'epidoxorubicin 60 mg + lipiodol 6 ml), repeated after 4 weeks (25) b) Symptomatic treatment (25)

22. Group Francophone of HCC [37]

96

Not evaluable

Median survival (weeks)

6.8 7.2

a) Chemo-L-TAE (lipiodol 10 ml + gelfoam + cisplatin 70 mg) every two months for a maximum of 4 courses (50) b) Symptomatic treatment (46)

* Five-month survival. b As indicated by authors, and/or inducing treatment withdrawal or modification of scheduling. Abbreviations: TAE = transartenal embolization; L-TAE = transarterial embolization + lipiodol; Chemo-TAE = transarterial embolization + chemotherapic agent; Chemo-L-TAE = transarterial embolization + lipiodol + chemotherapic agent.

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

* Mean survival. b As indicated by authors, and/or inducing treatment withdrawal or modification of scheduling.

129

1-Year survival

Severe adverse reaction

2-Year survival

Controls %

Odds ratio (95% CI)

Treated %

Controls %

(a) 16

(b)ll

1.6(0.3-7.6)

(a) 16

(b)ll

1.6(0.3-7.6)

(a) 22

(b) 5

3.8(0.6-25)

(a) 44

(b)33

2.4(0.5-10.9)

(a) 44

(b) 14

3.8(0.7-21)

V WBC ft GOT/GPT a) 17 8 b) 14 14

(a) 72

(b)54

2.1 (0.4-11.6)

(a) 40

(b) 9

5.7 (0.9-36)

0

(a) 48

(b) 9

7.4(1.8-30.3)

0

(a) 28 (b) 8

(c) 6 (c) 6

3.9(1.1-14) 1.3(0.2-9.8)

a) 0 b) ft Incidence cholestasis

(a) 51

(b)43

1.3(0.6-2.6)

1-Year survival

(a) 27

(b)29

Odds ratio (95% CI)

0.8(0.3-1.6)

Severe adverse reactionb %

2-Year survival

Treated %

Controls %

Odds ratio (95% CI)

Treated %

Controls %

Odds ratio (95% CI)

(a) 42 (a) 42

(b)21 (c)13

3 (0.8-10.8) 3.9(1.0-14)

(a) 25 (a) 25

(b)21 (c)12

1.3(0.3-5.7) 1.8(0.4-8.4)

(a) 24

(b)31

0.8(0.2-3)

(a) 73

(b)43

3.7(0.9-13)

(a) 20

(b) 20

1.0(0.2-3.9)

(a) 62

(b)43

2.1(0.9-4.6)

ft Creatinine 8 7

(a) + (b) Cholecystitis 5, acute gastric lesion 2

a) Acute renal failure 5, gastrointestinal bleeding 5 (a) 35

(a) 38

(b) 0

(b)26

9.8(1.9-50)

1.7(0.7-4)

Treatment related death %

a) Peptic ulcer 11 b) Hemorragic gastritis 6

a) Liver failure 60 Gastrointestinal bleeding 8 Cholecystitis 4 (JPLT2 V Serum sodium 2

Acute renal failure 5

Hemorragic gastritis 6

Liver failure 2

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

Treated %

Treatment related death%

130 Table 7. Patients characteristics, treatments and outcome rates of each trial. Embolization + chemotherapic agents studies. Mean age (range)

Treatment in each arm (number of patients for arm)

Authors [References]

Number of patients

23. Kawai[38]

289

- (39-83)

a) Chemo-TAE + doxrubicin 40 mg/mq (141) b) TAE (148)

24. Chang [39]

46

-(43-78)

a) Chemo-L-TAE (lipiodol 5-15 ml + gelfoam particles + cisplatin 50 mg) (22) b) L-TAE (lipiodol 5-15 ml + gelfoam particles) (24)

25. Ikeda[40]

135

59.5

a) Chemo-L-TAE (lipiodol + doxorubicin + mitomycin C + cis-diamminedichloro-platinum) (59) b) Chemo-L-TAE (lipiodol + doxorubicin + mitomycin Q (76)

26. Watanabe[41]

77

64(40-85)

a) Chemo-L-TAE (lipiodol + doxorubicin 40 mg/m 2 ) (38) b) Chemo-L-TAE (lipiodol + farmorubicin 60 mg/m 2 ) (39)

27. Ikeda[42]

40

61 (39-77)

a) Chemo-TAE (gelfoam + mitomycin C 4-20 mg or doxorubicin 10-30 mg) + tegafur 200 mg + uracil 448 mg/day orally (20) b) Chemo-TAE (gelfoam + mitomycin C 4-20 mg or doxorubicin 10-30 mg) (20)

Median survival (weeks)

22.7

28.2

Table 8. Patients characteristics, treatments and outcome rates of each trial. Embolization as adjuvant therapy. Authors [References]

Number of patients

Mean age (range)

Treatment in each arm (number of patients for arm)

Median survival (weeks)

28. Izumi[43]

50

62

a) Resection + arterial chemotherapy with or not TAE (gelfoam + doxorubicin + mitomycin C or hpiodol + doxorubicin + mitomycin Q (23) b) Resection (27)

102

29. Wu [44]

52

52

a) Resection + chemo-L-TAE (spongostan + lipiodol + doxorubicin) every 4-6 weeks (1-5 course) (24) b) Resection (28)

30. Lygidakis [451

40

63.5

a) Resection + pre- and post-operative immunochemotherapy (lipiodol + urographin + mitomycin C, carboplatin, farmorubicin, leukoverin, 5-FU, gamma-interferon 1° course; lipiodol + urographin + proleukin + immunokin 2° course before and after resection) (20) b) Resection (20)

85

* As indicated by authors, and/or inducing treatment withdrawal or modification of scheduling. Abbreviations: TAE = transarterial embolization; L-TAE = transarterial embolization + lipiodol; Chemo-TAE = transarterial embolization + chemotherapic agent; Chemo-L-TAE = transarterial embolization + lipiodol + chemotherapic agent. - Not reported.

so-called 'post-embolization syndrome' (transient abdominal pain and fever) occurred in 50%-60% of treated patients [37]. Thirty treated patients out of 50 in the French study developed liver failure (fatal in one patient) although only compensated patients in Child's A class were admitted. Liver failure or decompensation were reported less frequently (4%-30% in three studies [40, 42-44]. No data on post-treatment induced liver failure are reported in other studies. Less frequently

reported are acute cholecystitis, gastrointestinal bleeding, peptic ulcer. RCTs of immunotherapy Two RCTs of recombinant interferon-oc2 (IFN) were performed by the same authors (Table 9). In the first study, enrolling 75 patients [46], IFN was compared with doxorubicin; in the second [47] no active treatment was given to patients in the control group. In a third study [48]

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

* As indicated by authors, and/or inducing treatment withdrawal or modification of scheduling. Abbreviations: TAE = transarterial embolization, L-TAE = transarterial embolization + lipiodol; Chemo-TAE = transarterial embolization + chemotherapic agent; Chemo-L-TAE = transarterial embolization + lipiodol + chemotherapic agent.

131

1-Year survival

2-Year survival

Severe adverse reaction*

Treatment related death

Treated %

Controls %

Odds ratio (95% CI)

Treated %

Controls %

Odds ratio (95% CI)

(a) 74

(b)65

1.5(0.9-2.5)

(a) 52

(b)44

1.4(0.9-2.2)

Anemia II WBC V PLT (Not evaluable how many severe) a) 10 17 19 b) 22 25 14

(a) 52

(b)72

0.5(0.1-2.3)

(a) 26

(b) 39

0.6(0.2-2.1)

a) Gastric ulcer 4, esophageal varices bleeding 4 b) Cholecystitis 4

(a) 68

(b)64

1.2(0.6-2.4)

(a) 34

(b)47

0.6(0.3-1.1)

a) b) liver function deterioration 7, liver abscess 7

(a) 75

(b)70

1.3(0.5-3.3)

(a) 44

(b)44

1.0(0.4-2.6)

a) 0 b) 0

(a) 75

(b)95

0.2(0.04-1.2)

(a) 45

(b)60

0.6(0.2-1.9)

Decompensation a) 30 b) 5

Severe adverse reaction*

2-Year survival

Treated %

Controls %

Odds ratio (95% CI)

Treated %

Controls %

Odds ratio (95% CI)

(a) 87

(b)81

1.5(0.3-7.0)

(a) 69

(b)63

1.3 (0.4-4.3)

Hepatic injury a) 4

Biloma

(a) 75

(b) 98

0.2 (0.03-0.8)

(a) 45

(b)80

0.2(0.1-0.8)

Cholecystitis a) 4

UG bleeding 13

Not evaluable

activity of interferon-p was compared to a chemotherapic agent (menogaril). In these trials the IFN dosages were very high. In all studies survival was very poor suggesting that the patients were in a very advanced stage. In the first two studies a not significant longer survival in patients treated by IFN was found; in the third study, a slightly longer 1-year and a shorter 2-year survival was found for the patients treated by IFN-p. The pooled odds ratio of survival at 1 year was 2.2 (95% CI 0.94-5.2)

Treatment related death

Ascites 13 10 (post resection) 10 (post resection)

reaching statistical significance at 2 years (3.4, 95% CI 1.1-10.5) when there are few patients in both treated and untreated groups. Flu-like syndrome occurred in all of the treated patients. Other side effects appeared in 24/53 patients of the first study, in 31/35 cases of the second study requiring dose reduction or cessation in about 30% of the patients. Sixteen out of 31 treated patients in the Falkson's study had severe, life-threatening or lethal toxicities. In a fourth

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

1-Year survival

132 Table 9. Patients characteristics, treatments and outcome rates of each trial. Immunotherapy studies. Authors [References]

Number of patients

Mean age (range)

Treatment in each arm (number of patients for arm)

31. Lai [46] (cross-over for a subset of patients)

75

52 (28-73)

a) Recombinant alpha-interferon 18 x 106 IU/mq/day (25) + 50 x 106 IU/mq x 3/weekly (25) b) Doxorubicin 60 mg/mq (25)

32. Lai [47]

71

33. Falkson[48]

69

34. Kawata[49]

24

Median survival (weeks)

8.3 4.8

a) Recombinant alpha-interferon 50 x IU/mq x 3/weekly (35) b) No treatment (36)

14.5 7.5

Not evaluable

a) Beta-interferon 90 x 106 U i.v. 1-10 days every 3 weeks (21) b) Menogaril 200, 240 mg/mq i.v. every 4 weeks (29)

11 23

56

a) Resection + intra-arterial doxorubicin 13 mg/mq + recombinant interleukin-2 and lymphokine-activated killer cells (12) b) Resection + intra-arterial doxorubicin 13 mg/mq (12)

* As mdicated by authors, and/or inducing treatment withdrawal or modification of scheduling. b Relationship uncertain.

Authors [References]

Number of patients

Mean age (range)

Treatment in each arm (number of patients for arm)

35. Lai [50]

166

57(20-69)

a) No treatment (37) b) Hepatic dearterialization (33) c) Hepatic artery ligation and cannulation for chemotherapeutic agents infusion (30) d) Hepatic artery ligation and portal vein cannulation for chemotherapeutic agent infusion (29) e) Radiotherapy (37)

Median survival (weeks)

8.2 4.8 4.8 11.5 9.4

36. Order [51] (cross-over trial)

98

Not evaluable

a) 131 Antiferritin + doxorubicin 60 mg/mq + 5-FU 500 mg/mq (48) b) Doxorubicin + 5-FU 500 mg/mq (50)

Not evaluable Not evaluable

37. Raoul[52]

27

67

a) 131 I-Iodized oil (60 m CI) intrahepatic arthery repeated after 2, 5, 8 and 12 months (14) b) Tamoxifen (20-40 mg/day) and or symptomatic therapy (13)

24

' As indicated by authors and/or inducing treatment withdrawal or modification of scheduling. b Survival at 9 months. Abbreviation; TAE = lipiodol transcatheter arterial embolization.

study [49] immunotherapic agents were added to doxorubicin as adjuvant treatment post resection without evidence of benefit. Other studies Three studies were included in this subgroup (Table 10). In a 5-arm RCT [50], the one-year survival was exceedingly low or nil in each of the arms. The treatment could be changed during the trial period if no evidence of advantage was apparent. In the second study [51] comparing antiferritin added to doxorubicin and 5-FU with doxorubicin and 5-FU, survival was equivalent in the two

groups. Because of cross-over design and the unsatisfactory reporting of survival rate, results could not be tabulated. In the last study intra-arterial 131-I-iodized oil [52] was used in patients with portal vein thrombosis with apparent good results.

Discussion This review is limited to RCTs published as full papers since 1978. This date is important in the natural history of HCC as since then ultrasonography and improved tech-

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

Table 10. Patients characteristics, treatments and outcome rates of each trial. Other studies.

133

1-Year survival

2-Year survival

Severe adverse reaction*

Treatment related death

Cardiotoxicity + infection 25 Dementia and renal failure 4 b

Treated %

Controls %

Odds ratio (95% CI)

Treated %

Controls %

Odds ratio (95% CI)

(a) 15

(b) 0

4.6(0.5^(0)

(a) 13

(b) 0

4.9(0.8-29)

a) Cardiotoxicity 14, V blood cells 21, renal failure 7 b) (t blood cells 2, renal failure 2, dementia 2

(a) 23

(b)ll

2.3(0.7-7.8)

(a) 14

(b) 3

4.3(0.8-23)

a) V WBC 6, I] PLT 26 Mental deterioration 6, persistent fatigue 34

(a) 17

(b) 14

4.6(0.5-40)

(a) 2.5

(b) 5

0.7(0.1-12)

a) Neurologic, hematolo and genitourinary and other 39 b) Infection, mucocutaneous, neurologic, hematologic, other 30

(a) 100

(b)83

8(0.5-99)

(a)0 (a)0 (a)0 (a)0

(b)0 (c)0 (d)3 (e)0

Odds ratio (95% CI)

Treated

Controls

(b)0 b

Infection and hemorrhage 6

1 (0.1-8.2)

Severe adverse reaction11

Treatment related death

Odds ratio (95% Q )

b) Higher 7-day mortality than in other group

0.1 (0.0-5.3)

Not evaluable

(a)7 b

Infection 3

V Blood cells a) 58 b) 54 6.9 (0.9-99)

nique of AFP determination allowed the diagnosis of HCC before the terminal stage. We did not find RCTs of surgical resection or transplantation, presumably because they are assumed as definitely beneficial in a small population of selected patients. No RCTs of alcoholization are avaOable although promising results were obtained in comparative non randomized studies. No major breakthroughs have emerged in the area of treatment of hepatocellular carcinoma. Although this is generally acknowledged by clinical researchers in the field, it is less evident why this has occurred. This systematic review of 37 RCTs so far available in

Other 0 4

Allergic pneumonia to lipiodol'.

fully published form provides some insights on the problem even though it does not reliably indicates whether a particular treatment option is better than another. In general, most of the trials we reviewed had serious methodologic and clinical drawbacks (i.e., insufficient statistical power; exclusion of patients from the arm originally assigned, unplanned cross-over or association of treatments, comparison between almost identical treatments). In many trials the patients were not stratified according to prognostic indicators and the distribution of clinical aspects significantly associated with survival [53, 54] such as cirrhosis, age, performance status were not reported.

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

Controls %

83

2-Year survival

1-Year survival Treated %

83

0

134

TREATED WOflSE

TREATED BETTER

This is of special importance, since HCC has high prognostic variability, bound both to the si2e and aggressiveness of the tumor and to the stage and progressive tendency of the cirrhotic process. This, together with the evident lack of coherent coordination in the chronologic development of the hypotheses, explains why still a lot of uncertainty surrounds this area of clinical research, although the main reason for the current state of uncertainty is the lack of promising treatment. With regard to the treatments assessed in the trials here reviewed, there is no evidence of efficacy for chemotherapy both systemic or added to TAE or L-TAE; thus, its use is not justified. Tamoxifen is the only drug with some indication of benefit in patients with unresectable and advanced HCC while for embolization the wide variety of treatment modalities used does not allow any reliable conclusion. The high rate of side effects, above all decompensation, also in patients with initial good liver function, suggests great caution for its use. So far, no data justify the widespread, uncontrolled, use of TAE as adjuvant therapy. For immunotherapy, the results are insufficient to draw out conclusive indications. The trend towards positive results as expressed by the pooling OR are counter-balanced by the high rate of toxic events and by the finding of an increase of relative risk of dying found by Falkson in the treated patients after controlling for other covariates [48]. Finally, the efficacy and spectrum of application of alcoholization were not assessed in any prospective RCT, despite the extensive use of the procedure. As in any meta-analysis comprehensiveness in the identification of relevant studies is crucial to the validity of the results. In addition to a computerized MEDLINE

Acknowledgments The study was possible thanks to a generous grant from the Associazione Italiana per lo Studio del Cancro. The Authors wish to thank Mrs. Clara Profeta and Mr. Antonio Lino for preparation of the manuscript.

Appendix 1 Criteria used to analyze selected characteristics of methodology and in reporting of tbe reviewed studies a) Methodologic characteristics 1. Randomization method: - fully blinded randomization method such as telephone at a central office or by an equivalent arrangement made in single institution; - partially blinded randomization method such as sealed envelopes; - method without assurance of blinding (i.e., odd and even number, birthday, etc.); - method not evaluable because no information about this issue is reported. 2. Response evaluation: clinical (= 1), laboratory investigations or imaging (=2) parameters used to establish objective response. 3. Drop-out, withdrawals and patients with protocol deviation: - drop-out: patients lost to follow-up. Withdrawals patients withdrawn after randomization because not eligible and/or not evaluable. Protocol deviation: patients not treated according to randomization. 4. Analysis of drop-out, withdrawals and patients with protocol deviation: - 'intention to treat' analysis or with and without drop-out, withdrawals and patients with protocol deviation;

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

6. Randomized controlled trials for treatment of hepatocellular carcinoma. Interferon versus active or no active treatment. The results on 1-year survival are reported. A L'Abbe plot is shown on the left. In this plot the outcome rate in the experimental treatment group of each trial is shown on the horizontal axis and the outcome rate in the control group (with active or non active treatment) is shown on the vertical axis. Each dot represents the outcome rates in one trial. The plot shows the line of identity or equivalence between treatments. Dots below or above the identity line indicate that the treatment represented on the horizontal axis is respectively better or worse than the control represented on the vertical axis. The right side of the panel shows the odds ratios (log scale) with 95% CIs for the 1-year survival in treated patients compared with controls in each trial. Heterogeneity test: P = 0.1. TREATED: groups of patients given the experimental treatment. The numbers on the left of the bars, representing odds ratio, are referred to the studies listed on Tables 2, 3, 9.

search we perused all the reference lists of original and review articles and we feel it is unlikely that we have missed important published articles that might have changed our results. Of course we cannot rule out that a few studies may have been unreported because of publication bias. We however believe that this would not have substantially altered our conclusions. Another possible criticism to this review could be that we did not attempt at accounting in the analyses for by the large differences in the baseline characteristics of patients enrolled in different trials in different part of the world where the natural history of the disease is known to be different. While this is indeed a limitation inherent in using 'aggregate' rather than 'individual patient data1 in metaanalyses we feel that our analysis it would have been less helpful to the readers had we subdivided studies in further subgroups in addition to the ones already identified. In conclusion, when one attempts at a 'best evidence synthesis' of what is available on the treatment of HCC the emerging picture is gloomy. Several types of treatment (e.g. chemotherapy) were definitely non effective. The three treatments marginally (and uncertainly) effective (i.e. tamoxifen, IFN and embolization) should be further investigated by better, more carefully designed trials, whereas alcoholization should be assessed in prospective randomized controlled trials.

135 -

drop-out, withdrawals and patients with protocol deviation excluded from analysis; - not evaluable when no information was reported. 5. A priori estimate of sample size: - a priori estimate of the number of patients and the expected rate of events on which estimate was based; - the issue was not considered. b) Completeness of report ing

Appendix 2 In an overview expanded to June 1996, we retrieved only one RCT assessing percutaneous ethanol injection (PEI) combined with ChemoL-TAE, versus Chemo-L-TAE alone. Of this trial we report the same issues as those analyzed for the other studies included in the meta-analysis. Bartolozzi C, Lencioni R, Caramella D et al. Treatment of large HCC: Transcatheter arterial chemoembolization combined with percutaneous ethanol injection versus repeated transcatheter arterial chemoembolization. Radiology 1995; 197: 812-8. Period of conduct: 1991-1993. Period of reporting: 1995. Participation centers: < 5. Country of origin: Europe. Principal investigator specialty: Radiology. Type of journal: Radiology. Selected characteristics of design, execution and reporting Randomization method: not reported. Response evaluation: CTand MR imaging. Drop-outs, withdrawals and protocol deviations: not reported. Handling of drop-outs and withdrawals: not reported. A priori estimate of sample size: no. Selected aspects of the reporting of the studies reviewed Patient characteristics: age, gender, cirrhosis, Child's classification, tumor size and number.

Patient characteristics, treatments and outcome rates of each trial Number of patients: 53. Mean age (range): 65.7 (55-74). Treatment in each arm (number of patients for arm): a) Chemo-L-TAE (Lipiodol 5-20 ml + doxorubicin 20-70 mg + gelfoam), a single course + percutaneous ethanol injection - PEI (26 patients). b) Chemo-L-TAE (Lipiodol 5-20 ml + doxorubicin 20-70 mg + gelfoam), 2-5 courses (27 patients). 1-Year survival: a) 100%, b) 93%; odds ratio (95% CI) 7.4 (0.4-100). 2-Year survival: a) 87%, b) 70%; odds ratio (95% CI) 2.9 (0.8-11). No major treatment-related complication occurred. Abdominal pain and fever appeared, respectively, in 73% and 92% of cases after Chemo-L-TAE and in 69% and in 58% of cases after ChemoL-TAE + PEI. Chemical thrombosis of a segmental portal branch was observed after PEI. In conclusion, a statistically non-significant trend towards a better survival was found in patients treated with PEI + Chemo-L-TAE.

References 1. Rustgi VK. Epidemiology of hepatocellular carcinoma. Gastroenterol Clin North Am 1987; 16: 545-51. 2. London WT. Primary hepatocellular carcinoma-etiology, pathogenesis and prevention. Hum Pathol 1981; 12: 1085-97. 3. Lai CI, Gregory PB, Wu PC et al. Hepatocellular carcinoma in Chinese males and females, possible cause for the male predominance. Cancer 1987; 60:1107-10. 4. Muir C, Waterhouse J, Mack T et al. Cancer incidence in Five Continents. Vol V. Lyon: IARC Sci Publ 1987; 88. 5. Zamann SN, Johnson PJ, Williams R. Silent cirrhosis in patients with hepatocellular carcinoma. Cancer 1990; 65:1607-10. 6. Simonetti RG, Fiorello F, Politi F et al. Hepatocellular carcinoma a worldwide problem and the major risk factor. Dig Dis Sci 1991; 36: 962-72. 7. Okuda K. Hepatocellular carcinoma: Recent progress. Hepatology 1992; 15: 948-96. 8. Nerenstone SR, Ihde DC, Friedman MA. Clinical trials in primary hepatocellular carcinoma: Current status and future directions. Cancer Treat Rev 1988; 15:1-31. 9. Farinati F, De Maria N, Fornasiero A et al. Prospective controlled trial with antiestrogen drug tamoxifen in patients with unresectable hepatocellular carcinoma. Dig Dis Sci 1992; 37: 659-62. 10. Lok AS, Lai CL, Wu PC. Interferon therapy of chronic hepatitis B virus infection in Chinese. J Hepatol 1986; 3: S209-S215. 11. Okamura J, Kawai S, Ogawa M et al. Prospective and randomized clinical trial for the treatment of hepatocellular carcinoma: A comparison between L-TAE with farmorubicin and L-TAE with adriamycin (second cooperative study). Cancer Chemother Pharmacol 1992; 31 (Suppl): S20. 12. Kawai, Tani M, Okamura J et al. Prospective and randomized clinical trial for the treatment of hepatocellular carcinoma: A comparison between L-TAE with farmorubicin and L-TAE with adriamycin: Preliminary results (second cooperative study). Cancer Chemother Pharmacol 1994; 33 (Suppl): S97-S102. 13. Suto T, Fukuda S, Moriya N et al. Clinical study of biological response modifiers as maintenance therapy for hepatocellular carcinoma. Cancer Chemother Pharmacol 1994; 33 (Suppl): S145-S148. 14. Melia WM, Johnson PJ, Williams R. Controlled clinical trial of doxorubicin and tamoxifen versus doxorubicin alone in hepatocellular carcinoma. Cancer Treat Rep 1987; 71:1213-6. 15. Collins R, Yusuf S, Peto R. Overview of randomized trials of diuretics in pregnancy. BMJ 1985; 290:17-23.

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

6. Patients characteristics. Complete information about age, sex, presence/absence of cirrhosis, liver dysfunction*; performance status; tumour size; presence/absence of previous treatment * Liver dysfunction assessed by - Child-Pugh classification (which includes as items presence/ absence of ascites and of hepatic encephalopaty, serum level of albumin, bilirubin and protrombin time), or - Okuda prognostic stage, which include as items tumour size, presence/absence of ascites, levels of albumin and bilirubin); or assessed reporting; - some or all these criteria: ascites, hepatic encephalopaty, bilirubin, albumin, protrombin time. 7. Regimen description: - full description of the therapeutic regimen used, including dosage, schedule, toxicity indicators leading to dosage reduction; - incomplete information: only qualitative information or only information about dosage and/or schedule are reported. 8. Timing of event: - life-table or survival curve (actuarial or Kaplan-Meier) specifyng the number of patients at risk at every considered interval; - life-table or survival curve without patients at risk; - no life-table or survival curve. 9. Discussion of side-effects: - complete information: type of side effects and number of patients with side effects; - incomplete information: only partial information about type of side effects. - issue not discussed.

Regimen description: full information. Timing of events: survival curve without number of patients at risk. Discussion of side-effects: complete information. Note: large tumor (3-8 cm); no more than two daughter nodules; no infiltrative tumor or portal vein thrombosis, no extraepatic metastases.

136 38. Kawai S, Okamura J, Ogawa M et al. Prospective and randomized clinical trial for the treatment of hepatocellular carcinoma - a comparison of lipiodoltranscatheter arterial embolization with and without adriamycin (first cooperative study). Cancer Chemother Pharmacol 1992; 31 (Suppl 1): S1-S6. 39. Chang J-M, Tzeng WS et al. Transcatheter arterial embolization with or without cisplatin treatment of hepatocellular carcinoma Cancer 1994; 75: 2449-53. 40. Ikeda K, Inoue H, Yano T et al. Comparison of the anticancer effect of ADMOS alone and ADMOS with CDPP in the treatment of hepatocellular carcinoma by intra-arterial injection. Cancer Chemother Pharmacol 1992; 31 (Suppl): S65-S68. 41. Watanabe S, Nishioka M, Ohta N et al. Prospective and randomized controlled study of chemoembolization therapy in patients with advanced hepatocellular carcinoma. Cancer Chemother Pharmacol 1994; 33 (Suppl): S93-S96. 42. Ikeda K, Satoh S, Koida I et al. A prospective randomized evaluation of a compound of tegafur and uracil as an adjuvant chemotherapy for hepatocellular carcinoma treated with transcatheter arterial chemoembolization. Am J Clin Oncol 1995; 18: 204-10. 43. Izumi R, Shimizu K, IyobeTet al. Postoperative adjuvant arterial infusion of lipiodol containing anticancer drugs in patients with hepatocellular carcinoma. Hepatology 1994; 20: 295-301. 44. Wu CC, Ho YZ, Ho WL et al. Preoperative transcatheter arterial chemoembolization for resectable large hepatocellular carcinoma: A reappraisal. Br J Surg 1995; 82:122-6. 45. Lygidakis NJ, Potoulakis J, Konstantinidou AE, Spanos H. Hepatocellular carcinoma: Surgical resection versus surgical resection combined with pre- and post-operative locoregional immunotherapy-chemotherapy. A prospective randomized study. Anticancer Res 1995; 15: 543-50. 46. Lai C-L, Wu P-C, Lok AS-F et al. Recombinant a 2 interferon is superior to doxorubin for inoperable hepatocellular carcinoma: A prospective randomised trial. Br J Cancer 1989; 60: 928-33. 47. Lai C-L, Lau J-YN, Wu P-C et al. Recombinant interferon-a in inoperable hepatocellular carcinoma: A randomized controlled trial. Hepatology 1993; 17: 389-94. 48. Falkson G, Lipsitz S, Borden E et al. Hepatocellular carcinoma. An ECOG randomized phase II study of beta-interferon and menagoril. Am J Clin Oncol 1995; 18: 287-92. 49. Kawata A, Une Y, Hosokawa M et al. Adjuvant chemoimmunotherapy for hepatocellular carcinoma patients. Am J Clin Oncol 1995; 18: 257-62. 50. Lai E-CS, Choi TK, Tong SW et al. Treatment of unresectable hepatocellular carcinoma: Results of a randomized controlled trial. World J Surg 1986; 10: 501-9. 51. Order S, Pajak T, Leibel S et al. A randomized prospective trial comparing full dose chemotherapy to 1311 Antiferritin: An RTOG study. Int J Radiat Oncol Biol Phys 1991; 20: 953-63. 52. Raoul JL, Guyader D, Bretagne JF et al. Randomized controlled trial for hepatocellular carcinoma with portal vein thrombosis: Intra-arterial iodine-131-iodized oil versus medical support. J Nucl Med 1994; 35:1782-7. 53. Falkson G, Cnaan A, Schutt AJ et al. Prognostic factors for survival in hepatocellular carcinoma. Cancer Res 1988; 48: 7314-8. 54. Calvet X, Bruix J, Gines P et al. Prognostic factors of hepatocellular carcinoma in the west: A multivariate analysis in 206 patients. Hepatology 1990; 12: 753-60. Received 31 January 1996; accepted 3 October 1996.

Correspondence lo: Dr. R. G. Simonetti Divisione Medicina Interna Ospedale Cervello Via Trabucco 180 90146 Palermo, Italy

Downloaded from annonc.oxfordjournals.org by guest on July 13, 2011

16. L'Abbe KA, Detsky AS, O'Rourke K. Meta-analysis in clinical research. Ann Intern Med 1987; 107: 224-33. 17. Falkson G, Moertel CG, Lavin P et al. Chemotherapy studies in primary liver cancer. A prospective randomized clinical trial. Cancer 1978; 42: 2149-56. 18. Melia WM, Johnson PJ, Williams R. Induction of remission in hepatocellular carcinoma. A comparison of VP 16 with adriamycin. Cancer 1983; 51: 206-10. 19. Choi TK, Lee NW, Wong J. Chemotherapy for advanced hepatocellular carcinoma. Adriamycin vs. quadruple chemotherapy. Cancer 1984; 53: 401-5. 20. Falkson G, Maclntyre J, Moertel CG et al. Primary liver cancer. An eastern Cooperative Oncology group trial. Cancer 1984; 54: 970-7. 21. Falkson G, Maclntery J, Shutt AJ et al. Neocarzinostatin versus mAMSA or doxorubicin in hepatocellular carcinoma. J Clin Oncol 1984; 2: 581^t. 22. Bezwoda WR, Weaving A, Kew M, Derman DP. Combination chemotherapy of hepatocellular cancer. Oncology 1987; 44: 207-9. 23. Falkson G, Ryan LM, Lewis AJ et al. A random phase II study of mitoxantrone and cisplatin in patients with hepatocellular carcinoma. An ECOG study. Cancer 1987; 60: 2141-5. 24. Lai C-L, Wu P-C, Chan GC-B et al. Doxorubicin versus no antitumor therapy in inoperable hepatocellular carcinoma. A prospective randomized trial Cancer 1988; 62: 479-83. 25. Falkson G, Cnaan A, Simson IW et al. A randomized phase II study of acivicin and 4'deoxydoxorubicin in patients with hepatocellular carcinoma in an Eastern Cooperative Oncology Group Study. Am J Clin Oncol 1990; 13: 510-5. 26. Kajanti M, Pyrnonen S, Mantyla M, Rissanen P. Intrartenal and intravenous use of 4'epidoxombicin combined with 5-fluorouracil in primary hepatocellular carcinoma. A randomized comparison. Am J Clin Oncol 1992; 15: 37-40. 27. Farinati F, Salvagnini M, De Maria N et al. Unresectable hepatocellular carcinoma: A prospective controlled trial with tamoxifen. J Hepatol 1990; 11:297-301. 28. Uchino J, Une Y, Sato Y et al. Chemohormonal therapy of unresectable hepatocellular carcinoma. Am J Clin Oncol 1993, 16: 206-9. 29. Elba S, Giannuzzi V, Misciagna G, Manghisi OG. Randomized controlled trial of tamoxifen versus placebo in inoperable hepatocellular carcinoma. Ital J Gastroenterol 1994; 26: 66-8. 30. Martinez Cerezo FJ, Tomas A, Donoso L et al. Controlled trial of tamoxifen in patients with advanced hepatocellular carcmoma. J Hepatol 1994; 20: 702-6. 31. Manesis EK, Giannoulis G, Zoumboulis P et al. Treatment of hepatocellular carcinoma with combined suppression and inhibition of sex hormones: A randomized, controlled trial. Hepatology 1995; 21:1535-42. 32. Castells A, Bruix J, Bro C et al. Treatment of hepatocellular carcinoma with tamoxifen: A double-blind placebo-controlled trial in 120 patients. Gastroenterology 1995; 109: 917-22. 33. Lin D-Y, Liaw Y-F, Lee T-Y, Lai C-M. Hepatic arterial embolization in patients with unresectable hepatocellular carcinoma. A randomized controlled trial. Gastroenterology 1988; 94: 453-6. 34. Pelletier G, Roche A, Olivier I et al. A randomized trial of hepatic arterial chemoembolization in patients with unresectable hepatocellular carcinoma. J Hepatol 1990; 11:181-4. 35. Yoshikawa M, Sisho H, Ebara M et al. A randomized trial of intrahepatic arterial infusion of 4'epidoxorubicin with hpiodol versus 4'epidoxorubicin alone in the treatment of hepatocellular carcinoma. Cancer Chemother Pharmacol 1994; 33 (Suppl): S149S152. 36. Madden MV, Krige JEJ, Bailey S et al. Randomized trial of targeted chemotherapy with lipiodol and 5-epidoxonibicin compared with symptomatic treatment for hepatoma. Gut 1993; 34:1598-600. 37. Groupe d'Etude et de Traitment du Carcinome Hepatocellulaire. A comparison of lipiodol chemoembolization and conservative treatment for unresectable hepatocellular carcinoma. N Engl J Med 1995: 332:1256-61.