Our evidence mapping collected information from 44 SRs and 158 studies (most of them RCTs-92.4%) published between 2001 and 2017. The high number of studies reflects the increasing interest by users and physicians in assessing the potential therapeutic value of cannabis for several medical conditions.

We found that effectiveness and safety of medical cannabis has been evaluated in multiple medical conditions such as multiple sclerosis, movement disorders (e.g. Tourette Syndrome, Parkinson Disease), psychiatric conditions, Alzheimer disease, epilepsy, acute and chronic pain, cancer, neuropathic pain, symptoms related to cancer (e.g. emesis and anorexia related with chemotherapy), rheumatic disorders, HIV-related symptoms, glaucoma, and COPD.

Medical conditions addressed by these SRs have been previously identified by surveys about medical cannabis use [59–64]. One of the most representative surveys showed that cannabis was primarily used for back pain (11.9%), sleeping disorders (6.9%), depression (6.7%), pain resulting from injury or accidents (6.2%), and multiple sclerosis (4.1%) [61].

However, we noticed that the evidence for medical cannabis effects on these conditions is heterogeneous regarding the conclusions and the quality of the collected studies. Most of the conclusions extracted from SRs were classified as “probably beneficial” and “unclear”. Furthermore, for some comparisons, conclusions claimed by SRs were inconsistent and even contradictory. One example was the comparison of cannabis and cannabinoids with conventional antiemetics for chemotherapy-induced emesis [32, 34, 42, 44, 52, 55], where two SRs found a “probably beneficial” conclusions, while remaining four SRs claimed for an “unclear” conclusion or “no effect¨.

The evidence supporting the medical use of cannabinoids varies widely by clinical scenarios from high to low quality evidence. In fact, for some medical conditions, that we found in this evidence map, studies cannot reach firm conclusions, although RCTs have been conducted. While for other medical conditions, not showed in this evidence mapping, cannabis has been approved for use with only preliminary data (pre-clinical studies or observational studies) supporting the use, as is the case of hepatitis C, chronic renal failure, and posttraumatic stress disorders [65].

The research on health effects of cannabis and cannabinoids has been limited by regulatory reasons and policies in some countries, leaving patients and health care professionals without the evidence to make decisions regarding the use of cannabis and cannabinoids in local scenarios. Some barriers have been identified to conducting basic, clinical, and population health research on cannabis and cannabinoids, including regulations that restrict access to the cannabis products, funding limitations, and numerous methodological challenges [66].

In relation to funding, we found that most of the analyzed individual studies were sponsored by pharmaceuticals companies. Because of complexity of the research agenda in this field more funding sources and mechanism are needed to better understand the comprehensive health effects of cannabis.

There were also a number of methodologic limitations. The use of reliable placebos and well-selected active control compounds are needed for clinical trials, since the psychoactive and vasoactive effects of cannabis are a considerable challenge for effective blinding [66]. This limitation is important since 71% of the individual studies included in the SRs compared cannabis and cannabinoids against placebo.

Furthermore, restrictions on drug supply lead to the lack of standardization in potency or quantity of pharmacologically active constituents in cannabis products [66, 67]. This barrier leads to another limitation in conducting clinical trials reflecting in the wide variety of cannabis compounds assessed for a given medical condition.

Moreover, to get well-validated evidence it is necessary to have high-quality research. The quality of the SRs was moderate to high according to AMSTAR scores. However, the most frequent drawbacks were: failure to declare conflicts of interest, lack of likelihood of publication bias evaluation and absence of ‘a priori design’. Additionally, it’s important to state that beyond the quality of the SRs, its crucial to judge the quality of the individual primary studies to get a context of what evidence is telling us.

One strength of our evidence mapping is the use of a sensitive and comprehensive search strategy to localize the 44 SRs included as a source of information. We also used a broad definition of SR in order to obtain the largest number of documents. Additionally, this evidence mapping uses a friendly format to organize and classify research questions in PICO format. Findings are shown graphically to allow the identification of research needs, fields of controversy and the overall quality of the SRs included. Interventions were rated according to the conclusions stated by authors of the SRs. It is important to consider that this classification does not represent the effect of the interventions.

One limitation of this evidence mapping is that the quality of the studies included in each SR was not evaluated in addition to the quality of the SRs. Furthermore, as it is a characteristic of evidence mapping methodologies, we did not assess the quality of the evidence supporting the conclusions, which would have required the use of some complementary methodology such as GRADE. In addition, were describe the conclusions of the included studies according to how the authors declared them, however the direction of effects for each comparison should be deeply assessed by systematic reviews. Despite these limitations, this evidence mapping meets its objective of organizing and describing the available evidence as reported by the authors.