Potential risk of disease modifying therapies on neoplasm development and coadjutant factors in multiple sclerosis outpatients

Neoplasm development in Multiple Sclerosis (MS) patients treated with disease-modifying therapies (DMTs) has been widely discussed. The aim of this work is to determine neoplasm frequency, relationship with the prescription pattern of DMTs, and influence of the patients’ baseline characteristics. Data from 250 MS outpatients were collected during the period 1981–2019 from the medical records of the Neurology Service of the HUPM (Hospital Universitario Puerta del Mar)—in Southern Spain—and analysed using Cox models. Neoplasm prevalence was 24%, mainly located on the skin, with cancer prevalence as expected for MS (6.8%). Latency period from MS onset to neoplasm diagnosis was 10.4 ± 6.9 years (median 9.30 [0.9–30.5]). During the observation period β-IFN (70.4% of patients), glatiramer acetate (30.4%), natalizumab (16.8%), fingolimod (24.8%), dimethyl fumarate (24.0%), alemtuzumab (6.0%), and teriflunomide (4.8%) were administered as monotherapy. Change of pattern in step therapy was significantly different in cancer patients vs unaffected individuals (p = 0.011) (29.4% did not receive DMTs [p = 0.000]). Extended Cox model: Smoking (HR = 3.938, CI 95% 1.392–11.140, p = 0.010), being female (HR = 2.006, 1.070–3.760, p = 0.030), and age at MS diagnosis (AGE-DG) (HR = 1.036, 1.012–1.061, p = 0.004) were risk factors for neoplasm development. Secondary progressive MS (SPMS) phenotype (HR = 0.179, 0.042–0.764, p = 0.020) and treatment-time with IFN (HR = 0.923, 0.873–0.977, p = 0.006) or DMF (HR = 0.725, 0.507–1.036, p = 0.077) were protective factors. Tobacco and IFN lost their negative/positive influence as survival time increased. Cox PH model: Tobacco/AGE-DG interaction was a risk factor for cancer (HR = 1.099, 1.001–1.208, p = 0.049), followed by FLM treatment-time (HR = 1.219, 0.979–1.517). In conclusion, smoking, female sex, and AGE-DG were risk factors, and SPMS and IFN treatment-time were protective factors for neoplasm development; smoking/AGE-DG interaction was the main cancer risk factor.

In the literature, we found a disparity in the data for both neoplasm frequency in MS patients and the role of DMTs in the neoplasm development, from an increase in cancer-related deaths (1.9-fold) 6 to normal or decreased cancer prevalence, although the hazard risk depends on the type of tumour 7,8 . Increased cancer risk has been observed among patients treated with IFN, GA, NTB, and ALB 9,10 . The development of skin cancer is contemplated in the technical information about GA, FLM (basal cell and squamous cell skin cancers, Bowen's disease, melanoma, Kaposi's sarcoma), and ALB (papilloma) (available at http:// www. ema. europa. eu, European Medicines Agency, EMA, last accesed May 2021).
Therefore, the aim of this work is to investigate the therapeutic pattern followed and its relationship with the incidence of neoplasia in MS outpatients. We hypothesise that DMTs could constitute risk factors for the development of neoplasia. Patients' baseline characteristics were considered for their possible contribution to the development of neoplasms.

Methods
Patient recruitment. This retrospective study was based on outpatients from the Department of Neurology at the Hospital Universitario Puerta del Mar (HUPM), attended to in consultation from onset of MS between 1981 and 2019. According to the Spanish Statistical Office, the area served by HUPM had an average population of 220 thousand during this period (data available at http:// www. ine. es, Instituto Nacional de Estadística, INE). The data source was the digital medical record used by neurologists conducting a multidisciplinary team. Other independently-trained team members carried out data extraction and statistical analysis. Individuals previously suffering from a neoplasm or diagnosed with Clinically Isolated Syndrome were excluded. As a result, 314 consecutive cases were analysed, 250 of which fulfilled the inclusion criteria. The last case was diagnosed in September 2017, and the follow-up of patients ended on the last database registered individual (March 31, 2019).
Measures. The analysed variables were: (a) demographic and baseline characteristics: sex, age at MS onset (AGE-DG); disease phenotype; as possible contributing factors, neoplasm family history and smoking; (b) factors related to neoplasm development: presence and number of neoplasms; malignancy/benignity; tumour lineage; age at neoplasm onset; latency period from the start of DMTs until neoplasm appearance c) factors related to the medication received: type and number of drugs; use order; treatment time for each drug; length of treatment.
Statistical analysis. Categorical variables were expressed as number and percentage of observed data; numeric variables were represented as mean ± standard deviation and median (minimum-maximum). Association between categorical variables was contrasted by χ 2 test, or if these conditions were not verified, by Fisher's exact test. Quantitative variable comparisons were performed using the Student's t-test. Associations were considered significant when p < 0.05. For each patient, the observation period (survival time) started with MS diagnosis until neoplasm appearance or the end of the study (censored case).
Several covariate analyses were performed to estimate the role of DMTs and other possible contributing factors to neoplasm development. The Cox proportional hazard model was used to analyse the predictors associated with the hazard rate (HR), with a 95% confidence interval (CI). For assessing the proportional hazard (PH) for each predictor of interest, we also estimated p-values between the ranked survival and the residuals. When the predictors did not satisfy the PH assumption, an extended Cox model was used. For this purpose, we included time-dependent variables to measure the interaction factor with time exposure 11 . The data were processed using IBM SPSS Statistics 24 and Epidat 3.1 software.  ], mostly ranging from 20-40 years (62.4%, n = 156). Relapsing-remitting variant (RRMS) (83.2%, n = 208) was the predominant medical condition. Neoplasm family history was present in 10.4% (n = 26) and tobacco consumption in 44% (n = 110) of individuals. In particular, 24% (n = 60) developed some kind of neoplasm, alone or successively. Five patients suffered from a second (n = 4) or a third process (n = 1). Out of the sampled patients, 6.8% (n = 17, 28.3% of neoplastic patients) suffered malignancy and two of these individuals presented benign tumour. Mean age at tumour diagnosis was 46.2 ± 11.3 years (median 45.5 ) for neoplasm and 52.1 ± 8.4 for malignancy (median 53.0 [37-69]). Feminine gender, smoking, and family history were significantly present in neoplasm patients as compared to patients who did not present these characteristics (p < 0.05). Cancer patients were older at AGE-DG (39.8 ± 12.3 years, median 40 ] than the remaining individuals (33.58 ± 11.12, median 32.0 , n = 233) (mean (p = 0.027), median (p = 0.043)).
Treatment duration, overall and specifically for each drug, is recorded in Table 3. Median treatment length was around eight years in all groups (whole sample 8.  Figure 1 shows the drug prescription frequency in sequential therapy for the whole sample (detailed in Table 3) as well as the four groups studied (NEO: patients with neoplasm, NNEO: neoplasm-free patients; CANCER: patients with cancer; NCANCER: cancer-free patients). IFN was the main first choice in all groups (64.8%, n = 162). As a second option, a wider range of drugs was prescribed, with GA being used significantly more in neoplastic (23.3%, n = 14) as compared to "healthy" patients (12.1%, n = 23) (p = 0.024). Individuals who received IFN as a third option had previously received it as a first option. Only one patient was required to change to a fifth drug (in the following order: IFN, GA, FLM, DMF, ALB) during the observation period and developed malignancy.  A second Cox proportional HR model analysed the predictors for cancer-free time (Table 4). AGE-DG (HR = 1.056, CI95% 0.998-1.117, p = 0.057) and TAB (HR = 0.040, CI95% 0.001-2.027, p = 0.108) influenced malignancy. Increased age at diagnosis raised the risk, while smoking appeared to be a protective factor. However, when variable-interaction effect was estimated (HR = 1.099, CI95% 1.001-1.208, p = 0.049), AGE-DG, as a risk factor, increased in smokers and TAB increased cancer risk as AGE-DG increased. Thus, smoking becomes a risk

Discussion
Are detected frequency and malignancy type similar to that of the general population? In the European population 12 , breast (13.5% of all cancer cases), prostate (12.1%), and lung cancer (11.9%) represent 37.5% of all tumours, in a similar proportion to our data (35.3% of cancers, n = 7). However, in our sample, skin melanoma represented 11.8% (2 of 17), while, in Europe, it is the sixth most frequent cancer (3%); although non-melanoma skin cancer (particularly basal cell carcinomas) data were similar in sampled individuals (71.4% of cutaneous cancers) vs the general population (70-80%). We detected 21 types of benign neoplasms, many of them common in individuals over the age of 50: MEGUS in 1% > 50; Hürtle-cell adenoma of the thyroid in 0.5-1% of adults; hyperplastic colonic polyps in 30% of adults, 50% in the elderly 13 . Uterine myoma was the second most frequent neoplasm in the patients sampled, in consonance with myometrial tumours representing 20% of benign tumours in women (in our sample, 23.3% [n = 8] of 37 women suffered from benign processes) 14 .
Several studies suggested an increased risk of breast and central nervous system cancer, or benign neoplasm (meningioma, adenoma), but not especially skin cancer 7,9,15 . On the contrary, other works detected a decreased HR 16 : intense MS immune activity or immunomodulatory treatment has been hypothesised as an explanation 17 . In this sense, the Cox analysis showed that increased age at MS onset implied a greater risk of both neoplasm Table 3. Differences in therapy pattern between neoplasm-affected and unaffected patients.

Treatment length (years)
Mean ± SD 8.2 ± 5.  www.nature.com/scientificreports/ and malignancy. Would this be due to a "protective" effect of the MS treatment, or to senility itself? Untreated patients were older; both at MS onset and cancer diagnosis, and 27.8% developed cancer, a higher value than the 13.92% for a Spanish population ≥ 65 under multi-morbidity conditions 18 . Moreover, a population-based study (51% women; mean age 47.76 ± 10.99, 77% population > 24 years old) reported a prevalence of 3% 19 . Despite this, cancer prevalence in our sample (6.8%) was in the range of 2.6-7.3% recorded in the MS literature 10,16,20 . Neoplasm type and relative frequency were those expected in the general population, except for skin tumours (37.9%, 25 of 66 of neoplasms diagnosed), being the most prevalent. This is an expected side-effect of DMTs, recorded in the adverse reactions section of the Summary of Product Characteristics for some drugs such as FLM, as mentioned above, or GA: ≥ 1/100 to < 1/10 of GA-treated patients could develop benign skin neoplasms; ≥ 1/1000 to < 1/10 could develop skin cancer. Technical information on other more recently introduced active substances such as NTB (00') and ALB (10's) includes a recommendation for additional monitoring for any suspicious reaction detected (http:// www. ema. europa. eu), but this is rather directed at other health-compromising side-effects (hepatic, immunological, or haematological effects). However, NTB has been directly linked to melanoma replication, invasion, and migration via blockage of α4-integrin expressed in tumour cells, and to the development of melanoma in treated patients 21 .
Epidemiological studies on DMTs show data disparity 4 , although second-line immunosuppressants (azathioprine, cyclophosphamide, mitoxantrone) seem to involve a heightened risk for malignancy 22 . However, the relationship between immunotherapy and neoplasm development is inconclusive since some of these drugs are employed as coadjuvants or even anti-cancer drugs (ALB, cladribine, mitoxantrone) (http:// www. ema. europa. eu), or are under consideration (FLM, TER) as oncologic therapeutic options 7,23 . This implies their subjection by Medicines Agencies to the monitoring of the possible development of malignancy in treated MS patients. For example, in the safety analysis recorded in the technical information, the overall incidence of cancer was twofold higher in cladribine-treated patients compared to patients who received a placebo (http:// www. ema. europa. eu) 24 . It should be also highlighted that our patients will use them long-term.
Moreover, we found in our sample that TIFN and TDMF protected against neoplasm development, although INF and GA have been linked to cancer development 9,25-27 . However, a regressive effect of IFN has been casually observed in isolated neoplasms 28 , as well as a 32% lower mortality related to untreated patients 29 . Regarding DMF, which was recently licensed, a few studies have been published with a short follow-up, confirming its safe use (cancer was detected in only 0.9% of patients) 30 .
SPMS acted as a protective factor for neoplasm development, but large retrospectives studies did not find differences in phenotype contribution with respect to a decrease in cancer risk 32 . A possible explanation for our finding could be the distinct cytokine and adhesion molecule expression pattern of MS variants 33 . This should be extensively examined. Likewise, the influence of the patient's disability status on the neoplasm development could be considered. The latter is a limitation of our study and, although we did not include this variable like many other papers, some authors have found a low disease process, according to the Expanded Disability Status Scale (EDSS)(≤ 2), in the majority of patients with benign tumours 15 .
Finally, to understand the extent of neoplasm development in MS patients, genetic predisposal factors and lifestyle must be explored, although some works observe a lack of information in this regard 22 . Multivariate analysis revealed smoking to be the most significant risk predictor for neoplasm/cancer. Female sex is a risk factor; however, it should be noted that women predominated in our study (19.4%, 13 of 66 had gynaecological neoplasms). Family predisposition was a moderate risk predictor. In fact, a comprehensive study showed that lower cancer risk in MS patients did not coincide with a lower risk in their parents 34 .

Conclusions
In the patients studied, neoplasm prevalence was 23.4%, with a similar distribution of different types of tumours with respect to the general population, except for skin neoplasm (37.9% of occurring neoplasms). In our sample, 6.8% of patients suffered from cancer, in line with the data observed in other MS-focused studies. The extended Cox model identified smoking as the main risk factor for neoplasm development (HR = 3.938, CI 95% 1.392-11.140, p = 0.010), followed by the female gender (HR = 2.006, p = 0.030), and age at MS diagnosis (HR = 1.036, p = 0.04). SPMS (HR = 0.179, 0.042-0.764, p = 0.020) and treatment time with IFN (HR = 0.923, 0.873-0.977, p = 0.006) or DMF (HR = 0.725, p = 0.077) were protective factors. Tobacco and IFN treatment time lost their negative/positive influence as the result of an increase in survival time. The Cox regression model identified tobacco/AGE-DG a risk factor for cancer (HR = 1.099, CI95%1.001-1.208, p = 0.049), followed by FLM treatment time (HR = 1.219, p = 0.077).
In summary, genetic factor, lifestyle, the inflammatory profile of MS, drug type, and clinical practice interact in a complex manner. The last drugs introduced would require more clinical experience. Perhaps, exposure time of these risk factors should be taken into account, given the long-term nature of the disease.