Tissue damage associated with antineoplastic extravasation is an adverse event related with intravenous drug administration, whose risk varies greatly. Severity may range from mere erythema to tissue necrosis that even requires surgical debridement1. The incidence of extravasations reported is between 0.10% and 6.00% in peripheral venous administrations. In central venous administrations it is lower (0.26% to 4.7%) but more severe2. These values have come down in recent years, and currently range between 0.05% and 5.00%, thanks to improvements in training and to the implementation of strategies to minimise the incidence of extravasations3. Prevention is the best way to manage extravasations, but immediate action is necessary when they occur, in order to avoid or minimise associated tissue damage, which can compromise the clinical outcome of patients due to potential delays in their antineoplastic regimen, and also affect their safety and their quality of life.

Hospitals have protocols for the management of extravasations. These may be institutional or regional, and are approved by multidisciplinary consensus, serving as a guide for implementation on the basis of the tissue-damage classification of the involved drugs. It is therefore important to be familiar with the risk of tissue damage, with a view to making decisions relating to protocols and clinical practice. However, classifications vary depending on the source of reference, and there is even variability in the categories used to classify antineoplastics in terms of their potential for causing tissue damage4–6, which is usually defined on the basis of maximum described toxicity levels. Depending on the probability of extravasation-related damage, antineoplastics are classified in three categories (vesicant, irritant and non-irritant) or five (vesicant, exfoliant, irritant, inflammatory, and neutral). The first of these two classifications is the most frequently used in the literature. Regarding potential for tissue aggression, and although this depends mainly on the type of cell-damage mechanism involved, a series of variables come into play: concentration, extravasated volume, and pH and osmolarity, among others7,8. This adds further difficulty to antineoplastic classification.

Information on the effect and evolution of extravasations is largely based on cases series or case reports, with the exception of two small clinical trials involving dexrazoxane9. Little information is available regarding tissue damage in the technical description sheets of drugs. In some instances, data are derived from animal studies, which cannot be extrapolated to humans, while in other cases they are taken from outdated studies based on techniques and administration methods that are no longer in current use. In view of the scarcity of evidence and the degree of variability among the classifications proposed in the different guidelines, the working group on extravasations of the Spanish Oncology Pharmacy Group (GEDEFO), belonging to the Spanish Society of Hospital Pharmacists (SEFH), decided to apply the Delphi methodology to arrive at a consensus with regard to the tissue damage classification of antineoplastic drugs.

The Delphi method is an expert consensus technique that has been used in the field of healthcare when scientific evidence is either not available, scarce, or poor in quality; in such scenarios, the experience and accuracy of group judgements is considered to be superior to those of individual approaches, since they bring together the knowledge and experience of all group members10,11. In addition, the method preserves voting anonymity among the evaluators, thereby guaranteeing independent decisions on the part of group members. Other advantages include the fact that all panellists have equal opportunities when contributing to the final decision, and that the procedure is decentralised, with information and decisions being aggregated by summarising the results12. This methodology has been used in the hospital pharmacy practice to design satisfaction surveys, set up the taxonomy for pharmaceutical interventions, select training goals, or define quality indicators, among other objectives13–16.

The goal of the present paper is to establish a consensus of experts, by using the Delphi method, that can be used to classify the tissue-damaging potential of antineoplastic drugs, in order to facilitate the decision-making process in the event of extravasations.

The modified Delphi method was applied to classify antineoplastic drugs according to their potential for extravasation-related tissue damage, based on a consensus of experts. The panel of expert evaluators included seven pharmacists with oncological training, who were members of the GEDEFO's Extravasation Working Group and had previous expertise and publications dealing with this issue. Other of the group members served as coordinator. The study was carried out from October to December of 2019.

For the classification of antineoplastic drug-induced tissue damage, eight reference documents were selected, following preliminary research by the working group, which involved the examination of clinical extravasation guidelines published by scientific societies or institutions, routinely used databases, and reference papers dealing with extravasations4–6,17–21. In addition, and in order to identify emerging risks or new tissue-damage evidence, the scientific bibliographical search used in the Monograph for Prevention and Treatment in cases of antineoplastic extravasations was reviewed on PubMed database (search terms: “Antineoplastic Agents” [Mesh]) AND “Extravasation of Diagnostic and Therapeutic Materials” [Majr]. Filters: population, humans; language: English, French, Spanish. Advanced filters: title “extravasation”. Period: 01/01/2010–30/09/2019)22.

The present consensus has adopted the standard classification of tissue toxicity (vesicant, irritant, non-irritant) since it is the most frequently used in published studies and guidelines, but includes the modification proposed by Conde-Estévez et al.18, which distinguishes between high-risk and low-risk irritant antineoplastics, because of the differences in the management to prevent potential patient morbidity. Antineoplastics have therefore been classified in four categories, depending on the probability of tissue damage in the event of extravasation: vesicant (V); high risk irritant (HRI); low risk irritant (LRI); and non-irritant (NI). Table 1 summarises the definition for each category. Radionuclide conjugated monoclonal antibodies have been excluded from the classification, since tissue damage is in their case mainly associated with radiation, and their management differs from that of the rest of antineoplastics, hormone therapy, drugs that are currently not available in the therapeutic armamentarium, such as mechlorethamine, and investigational antineoplastic agents. Non-conjugated monoclonal antibodies were considered as a single group in terms of the classification.

The coordinator prepared an initial table indicating the category of tissue damage assigned to each antineoplastic on the basis of the eight reference documents, and this was sent by e-mail to each of the expert panel members. Based on the information included in that document, and on their own professional experience, evaluators issued their classifications (first round) which were returned to the coordinator. The latter examined the degree of agreement (percentage of evaluators in the most voted category) and the comments of group members, and issued a new document indicating the category with the highest score for each antineoplastic and including comments that were proposed anonymously. This new information was sent individually to each expert, for re-evaluation or confirmation of his/her previous classification (second round).

If agreement in the classification of any given antineoplastic was lower that 70% (5/7 members of the panel) at the end of the second round, the case was appraised individually, on the basis of additional information research and non-anonymous group discussion, and a joint decision was reached.

Finally, the definitive classification was sent to all group members, for individual approval. This classification has been included in the Monograph for Prevention and Treatment in cases of antineoplastic extravasations that was prepared by the group22.

In each round, the median of the degree of consensus with the interquartile range (IQR25–75) was examined, and for each category of tissue-damage risk the rate for agreement was analysed. The percentage of antineoplastics with a consensus of above 85%, and of 100%, was calculated. If agreement was above 80%, consensus was considered to be strong11,23. A separate analysis of the consensus achieved was performed considering only the drugs with discrepancies between their classification in the different reference documents and the rate for agreement reached.

Statistical work was performed using IBM SPSS v23.0 software (IBM SPSS Statistics for Windows, IBM Corporation, Armonk, NY).

After applying the defined exclusion criteria, 71 antineoplastics were selected to evaluate their tissue-damage risk, non-conjugated monoclonal antibodies were considered a single agent. On the basis of the search strategy that was employed, no additional literature was found to provide new information regarding the tissue-damaging potential of the antineoplastics that were being classified. Non-conjugated monoclonal antibodies were classified, by unanimous agreement, as NI.

In the first round, the median rate of consensus was 100.0% (IQR25–75: 71.4–100.0%). Of the total amount of evaluated antineoplastics, 18 drugs were classified as vesicant, 13 as HRI, 14 as LRI, and 26 as NI. In 66.7% (48/71) of the antineoplastics, the rate for agreement was 85.7% or above. Consensus was 100.0% in 52.1% (37/71) of the antineoplastics.

Thirty antineoplastics (42.3%) exhibited tissue-damage classification variability in the different reference documents; in these, the median rate for agreement was 71.4% (IQR25–75: 57.1–87.7%). Of the 30 drugs, 3 were classified as vesicant, 10 as HRI, 8 as LRI, and 9 as NI. In 40.0% of them (12/30), the rate of consensus was 85.7% or above.

In the second round, the median rate of consensus was 100.0% (IQR25–75: 85.7–100.0%). None of the antineoplastics modified their classification in the second round, but the percentage of agreement increased. In 85.9% (61/71) of the analysed antineoplastics, the degree of consensus among the experts was 85.7% or above; in 71.8% (51/71) a consensus of 100.0% was reached.

The median for rate of consensus of the 30 antineoplastics whose classification varied in the reference documents was 100.0% (IQR25–75: 85.7–100.0%) in the second round. None of the antineoplastics changed their classification. In 76.7% (23/30) of the antineoplastics, the rate for agreement was 85.7% or above.

Tables 2 and 3 show the rate for agreement in the classification of tissue-damage risk between the first and second round for each category of tissue damage. The agreement percentage is lower for both categories of irritant drugs.

At the end of the second round, only four antineoplastics exhibited a degree of consensus that was below 70%: busulfan, carmustine, etoposide phosphate, and treosulfan. They were subject to non-anonymous group discussion. For carmustine and busulfan a consensus of 100.0% was reached; consensus for treosulfan was 85.7%. In the case of etoposide phosphate, the level of agreement remained below 70.0%.

The final consensus on the classification of the antineoplastics, according to their potential for causing tissue damage in the event of extravasation, is summarised in table 4: 17 were classified as vesicants; 15 as HRI; 13 as LRI; and 26 as NI. In 90.1% of the antineoplastics (64/71) the rate for agreement of the reviewers, regarding tissue damage classification, was 85.7% or above. In 74.6% of the drugs (53/71) consensus was 100.0%.

If we refer these values to the analysis of the 30 antineoplastics which exhibited variability in their classification in the literature, we find that in 86.7% of them (26/30) the rate of consensus was 85.7% or above. In 60.0% (18/30) it was 100.0%.

The Delphi method is a consensus methodology that allows for decision-making in scenarios of uncertainty, and is widely used in healthcare in general and pharmacy in particular. In that context, the opinions of the expert panel members may help professionals to make decisions and generate new ideas, and the method has been used to develop guidelines, design guides, and define indicators or clinically significant interactions11,23–26.

There is no standard method for calculating the number of experts needed in order to apply the Delphi method; this depends on the objective of the study and the sources available10. In published studies, the number of experts varies, and usually includes about 12 to 1512,14,25, although in general it is considered that the minimum required number would be 7, and the maximum number would be 3010,13,16. The number of experts in the present study's panel met that criterion.

Regarding the number of evaluations required for reaching consensus, most studies have used two waves or two rounds, although there are differences between the published papers; in general, as a function of the degree of consensus achieved, the Delphi process is discontinued when the predefined consensus level have been reached12,14,25,27. In the present work, after the second round, a degree of consensus of above 85% was reached for more than 80% of the antineoplastics. By category of tissue aggression, the rate of consensus achieved was lower for irritant antineoplastics (both HRI and LRI); this is related to the subclassification introduced to take account of the differences in management of extravasations. Mader recommends the use of the standard classification, since it is still the most widely employed, but points out that other authors propose additional categories (five grades) in an attempt to improve both the classification and the management of extravasations6,8, which was also the objective of the present study.

The rate for agreement that is considered to be high level is different for the various published studies, ranging from 81.8% to 90.0%12. Some authors propose a minimum agreement rate of 80% as a consensus thres-hold24. When numerical values are used (scoring methods such as the Likert scale, for instance) a median of 77.77% or above is considered adequate11. When at least one third of the evaluators differ from the rest, no agreement or consensus is considered to have been reached11. If we apply the four categories of evidence and consensus of the National Comprehensive Cancer Network (NCCN) criteria (from a higher to a lower level of evidence and consensus: 1, 2A, 2B and 3), the recommendations of the present expert panel would fit into category 2A for most of the antineoplastics, since they are based upon lower-level evidence, but with a uniform classification consensus28.

Although expert consensus is considered to be the lowest level on the pyramid of scientific evidence, the strength of an expert consensus depends not only on the input available to the experts (systematic reviews, individual experiments, personal experience and qualitative studies) but also on the methodology used to achieve consensus26. In spite of its limitations, the present consensus is a starting point that establishes a practical basis for recommendations in the management of extravasations of the different antineoplastics, with the aim of reducing variability in patient care.

It is important to point out that classifications have traditionally been based on the highest grade of toxicity reported, irrespective of its actual frequency, even when anecdotal, or influenced by other factors related to the patient or the procedure, which could condition the final level of damage and the patient outcomes. A publication bias cannot be ruled out in cases whose consequences are more severe in the event of extravasation. Such bias may affect the classification by overestimating risk.

The leading factor in tissue damage is the mechanism of action, at cell or molecular level, of each cytotoxic; however, other aspects also contribute to tissue damage, and these include concentration, extravasated volume, ph (≤ 5 or ≥ 9), osmolarity (> 500 mOsm/L), excipients (Polysorbate 80 or Cremophor EL® increase irritating properties), formulation (liposomes), location, and the time that elapses from the moment of extravasation until the latter is detected7. Although these variables are not always described in the published case reports, they may affect the final outcome, and therefore influence the classification and the variability found in the literature.

Thus, the distinction between a vesicant or an irritant antineoplastic is not absolute, and depends on other factors: concentration, for example, is a variable that in most guidelines causes cisplatin to be considered HRI if it equals 0.4 mg/mL or above, or vesicant if it is below that value. It has recently been suggested that the administration of diluted vincristine reduces its tissue-damaging potential as compared to the concentrated preparation (1 mg/mL)29. Liposomal presentations are considered to be less likely to cause lesions that are compatible with vesicant damage, although no distinction is normally made between the pegylated and non-pegylated formulations. The expert panel considers that liposomal presentations are associated with a lower risk of tissue damage, although little information and published experience are available. Information is even scarcer with regard to antineoplastics that have been added to clinical practice in recent times. If no data are available regarding the tissue-aggression risk of a given drug, it is recommended to classify it as an irritant if it causes phlebitis and/or sclerosis at the injection site or along the vein22. The expert panel recommends adopting the tissue damage classification when a new antineoplastic drug is approved for use at each institution, and updating it as new information becomes available.

The only drug whose final agreement was below 70% was etoposide phosphate, which is not included in many of the reviewed reference documents (or not differentiated from standard etoposide), probably because of the greater weight of the evaluators’ personal experience regarding the classification of the reference documents, since the excipient of the standard etoposide is considered to contribute to potential tissue damage.

The present tissue-risk classification consensus provides significant differences as compared to previous classifications in the reference documents: it updates the included drugs, divides drugs into HRI and LRI categories, on the basis of their different management approaches, and uses an explicit consensus methodology that is widely employed in the field of healthcare. Thus, by applying the same exclusion criteria of the present study to the drugs that were classified in the reviewed reference documents, the tissue-damage risk of 26 drugs that were not included in the Micromedex database21, and 30 drugs not included in Uptodate20, has been classified, as well as 15 additional drugs that were not included in the most recent or permanently updated guidelines5,19. The classification that has been introduced distinguishes potential severity in the event of extravasation of an irritant antineoplastic agent, with a view to guiding management and subsequent follow-up18. The greatest classification differences in the various guidelines apply to irritant drugs, with regard to which different terminologies are even used; this is the case, for example, in irritants with vesicant properties, inflammatory agents and exfoliantdrugs6. Lastly, no specifications regarding the methods used for risk classification are included in the reviewed documents.

In conclusion, in this area of scarce evidence and given the variability in the tissue-damage risk reported in the literature, the Delphi expert consensus methodology facilitates decision making when reviewing the different reference sources and allows the experience of the panellists to be included. In the final classification, for approximately 90% of the antineoplastics, the rate of consensus of the expert panel was above 85%; for 74% of the antineoplastics, consensus was 100%, and offered a solid ground for management decisions.

No funding.

No conflict of interest.