1. Introduction

Limb trauma represents the most prevalent survivable major combat injury. Collectively, for all belligerent nations, the First World War yielded the largest military amputee cohort in history. In excess of 700,000 British soldiers received limb wounds during the conflict, resulting in an estimated total of more than 41,000 surviving young male amputees in the United Kingdom alone.5 The pattern and scale of limb wounds sustained during the First World War was unlike anything seen in prior conflicts. The combination of penetrating injuries from high-velocity projectiles, consequential major tissue disruption, and a high risk of infection in contaminated wounds (particularly by anaerobic microbes in the preantibiotic era) predicated a low threshold for the consideration of early life-saving amputation. Without effective vaccinations or antibiotics, debridement and early surgical intervention were often the most appropriate surgical tactics. Early amputation was further utilised due to the need to process unprecedented numbers of casualties through the evacuation pathway and avoid the time consuming and often ultimately futile job of limb reconstruction. Of all injuries amongst British soldiers in the First World War that were not immediately fatal, an overall proportion of 13% resulted in amputation.20 These numbers are unequalled by any subsequent conflict. For comparison, the Second World War led to approximately 12,000 British amputee veterans and the most recent Afghanistan conflict resulted in 302 UK service personnel undergoing one or more traumatic or surgical amputations between 2001 and 2020 (a total of 0.2% of the 150,610 British personnel who served in Afghanistan).50,51

Significant residual limb pain affects up to 85% of today's military amputees, and phantom limb is reported by at least 59% of all military amputees.1,26,37 If the same were true for the First World War cohort, then 35,000 British amputee veterans may have experienced chronic pain as a result of their amputation. Chronic pain has significant negative effect on the quality of life.53 Yet there has been no detailed analysis reporting the effect of conflict-related chronic postamputation pain on veterans' long-term health and quality of life, on resulting years lost to disability, in the evolution of medical attitudes, the clinical assessment, or the management for postamputation pain.68

This systematic review is part of a wider programme examining this topic, including conventional historiography and analysis of archival material, notably First World War veterans' pension files, and a prospective long-term follow-up of British injured veterans from recent conflicts.2,3,24 Although articles from medical journals of this period are often cited by historians, the cited sources are usually selected without a systematic approach, with the inherent risk of bias that such an approach risks.

It has been suggested that pain is often absent in published medical accounts due to a professional reluctance to discuss a condition that could not be surgically resolved. Edwards et al. considered that “because there was not potential for surgical resolution and it did not affect tissue viability for prosthetics, it [pain] was marginalised in medical discussion of amputation during World War One and in the period of reflection afterwards.”26 Establishing the extent to which this knowledge and interest was disseminated amongst medical professionals was one of the key objectives of this review. The goals of this systematic review were therefore to explore, using a systematic search of professional medical journals:

(1) the professional medical conversation on the aetiology (in terms of mechanistic descriptor), contemporary treatment, clinical presentation, and assessment of chronic postamputation pain in veterans injured on active service during the First World War; (2) the extent to which these developed over the lifetimes of these veterans.

Secondary aims were to identify when chronic postamputation pain (including phantom and residual limb, as defined in Table 1) became recognised as a potential disability either in its own right or as a contributing factor, when concepts of interdisciplinary or multidisciplinary treatment for chronic postamputation pain emerged, and the extent to which emergent medical specialties (eg, neurology, anaesthetics, psychiatry, orthopaedics, pain medicine) started to contribute to pain management.

Table 1 - Definitions of postamputation pain as outlined by Edwards et al. and the International Association for the Study of Pain.25,33,34 Residual limb pain Spontaneous (continuous or paroxysmal) or evoked pain perceived as originating in the residual limb including the stump; pain unrelated to amputation, eg, other injuries, such as damage of the nerves above the level of amputation Stump pain Spontaneous (continuous or paroxysmal) or evoked pain in the amputation stump, includes neuroma, muscle, and bone stump as pain sources Phantom limb pain Spontaneous (continuous or paroxysmal) or evoked pain perceived as arising in the missing limb Phantom limb sensation Any sensation of the missing limb including pain Neuropathic pain Pain caused by a lesion of disease of the somatosensory nervous system
2. Methods

The review protocol was prospectively registered with the Open Science Framework on May 4, 2020 at osf.io/cr5ab (DOI: 10.17605/OSF.IO/CR5AB).

2.1. Search criteria

A search of the 2 major professional medical journals of the 20th century in the United Kingdom, The Lancet and the British Medical Journal, for the full years 1914 to 1985 was undertaken. This time window is in line with the medical pension files for the First World War veterans held by the UK's National Archives at Kew catalogued in the file series and referred to as “PIN 26.”24 The search was intended to retrieve all articles that described the prevalence, assessment, or pathophysiological pain mechanisms of postamputation pain sustained by veterans of the First World War, as well as clinical descriptions, case histories, and treatments. The search was not limited to specific types of text, with all original research studies, reviews, editorials, conference reports, and correspondence included. As not all past issues of the journals were available on standard medical databases (eg, PubMed), The Lancet was searched through its own archives with texts retrieved through ScienceDirect, whereas JSTOR was used to search and retrieve texts from the British Medical Journal. Searches were conducted on June 3 to 4, 2020 (see Appendix, available online at https://links.lww.com/PR9/A204).

Table 2 - Inclusion criteria for text screening. Population First World War veterans with limb injury and amputation sustained on active service Intervention Any treatments intended to alleviate postamputation pain Comparison A range of methods were employed but no randomised control trials. Therefore, there are often no comparisons. However, comparisons were extracted where the author has described them Outcome Often no recorded outcomes or imprecise descriptions. Therefore, anything reported by the author as an outcome was extracted
2.2. Selection criteria

Included reports covered participants who were military veterans identified as having sustained a limb injury whilst on active service during the First World War (Table 2). In line with the British government's definition, “veterans” were all those who served one day or more in the armed forces.51 In the registered protocol, we had intended to include veterans with chronic pain due to amputation and peripheral nerve injury. However, due to the nature and number of retrieved texts, a protocol amendment decision was made by the first and senior author to focus on postamputation pain only. Texts on peripheral nerve injury were thus excluded during full-text screening.

Reports regarding veterans from all subsequent conflicts and civilians were excluded. The participant's injury and date range were the only specific inclusion criteria for the study because it was intended to explore the professional medical “conversation” around these conditions, their aetiologies, and their treatments as broadly and inclusively as possible. Limiting the search with specific inclusion and exclusion criteria based on particular interventions or study designs would have been overly restrictive and could create selection bias. Due to the age of the texts, it was not possible to seek a greater level of detail through contacting authors or searching unpublished sources.

2.3. Study selection and data extraction

Screening was performed by one author (S.D.S.) with inclusion conflicts and uncertainties resolved by conversation with the senior author (A.S.C.R.). Data extraction was performed by one author (S.D.S.).

Deduplication, title, and full-text screening were performed using Covidence (Veritas Health Innovation, Melbourne, Australia). It had been intended to screen studies by title and abstract before a full-text screening. However, due to the age of the texts, the method of digitisation and the wide range of article types identified in the search, none of the retrieved texts included an abstract, so initial screening had to be performed based on title alone. As the search covered almost a century of publication, there was no standardisation of format or terminology within the text, and data were extracted only in a qualitative manner. Extracted data were recorded in Microsoft Excel 365 and included details on patients, wounds sustained, treatment, and assumed mechanistic descriptors of pain. Results were exported into the qualitative analysis software, NVivo 12 (QSR International, Doncaster, Australia).

This study was conducted using archival material and was intended to explore the professional medical conversation across the 20th century, with a particular focus on the dynamics of how clinicians discussed and shared these ideas. Hence, a meta-analysis or a risk of bias assessment would have been inappropriate and was not conducted. Therefore, the output is in the form of a descriptive synthesis, structured around the aetiology and mechanistic descriptors of chronic postamputation pain and treatments strategies applied.

2.4. Terminology

We attempted to employ a classification of pain in our findings, using the terminology of the texts' authors. Doing so on a consistent and widely agreed basis proved problematic as definitions and criteria for diagnoses changed over the period, and attempts to formalise classification of pain are a recent phenomenon, which postdate our search period. Chapters in the classification of chronic pain were only included for the first time in the 11th iteration of the International Classification of Diseases (ICD-11) in 2019.68

As a result, postamputation pain was defined with the definitions in Table 1, taken from Edwards et al.26 Neuropathic pain was defined, as far as possible given the historic nature of the text, in broad alignment with current era definition published by the International Association for the Study of Pain, which is consistent with the recently published inaugural pain chapters in ICD-11.33–35,60 The historical and diverse nature of the reports resulted in a necessary degree of diagnostic imprecision, with reviewers exercising a degree of pragmatism. We based our classification largely on the history of nerve injury (inherently a feature of limb amputation), symptom descriptors, and the authors' use of contemporary terms such as “neuritis,” “neuralgia,” and “causalgia.” This would be broadly consistent with the “possible” level of diagnostic certainty in modern algorithms, the use of which we have previously reported in a historical context.29,57

3. Results 3.1. Article types

After deduplication, 8981 reports were assessed for inclusion. Screening by title excluded 4600 texts, and full-text screening removed a further 4280. One hundred one texts were thus included in the final data set (Fig. 1). Although the search covered the years 1914 to 1985, the most recent relevant text to be included in the final data set was published in 1956. That no relevant texts were retrieved from subsequent decades probably reflects the decreasing number of living veterans of First World War. Seventy-one of the texts were retrieved from the British Medical Journal and 30 from The Lancet. As anticipated, the discussion around conflict-related chronic pain peaked during the First World War, and the number of texts declined rapidly after 1919, with only a small increase around the Second World War (Fig. 2).

Figure 1.:

Retrieved texts for The Lancet and The British Medical Journal, with reasons for exclusion.

Figure 2.:

Total retrieved publications by decade for The Lancet and the British Medical Journal.

Retrieved texts categorised by article type are shown in Table 3. Conference summaries accounted for the largest type at 22 (22%). Narrative reviews and articles reflecting on previous cases, categorised as “in my experience,” each totalled 19 (19%). The remainder were 18 case reports (18%), 13 book reviews (13%), and 6 “other” (6%), the majority of which was correspondence. No randomised controlled trials testing an intervention were identified.

Table 3 - Categories of texts retrieved. Category Total 1914–19 1920–29 1930–39 1940–45 1946–49 1950–59 Book review 13 8 1 2 1 1 — Case report series 16 6 5 1 — 2 2 Conference summary 22 15 3 2 2 — — “In my experience” 19 10 2 1 1 4 1 Narrative review 19 13 2 — 4 — — Other 6 5 — — 1 — — Question and answer 2 — — 1 — — 1 Single case report 2 — 1 — — 1 —

In total, 50 articles referenced other texts with a total of 170 references: 26 (15%) of which were in another issue of the same journal, suggesting a response to an article or correspondence, and 53 (31%) were in another professional medical journal. A further 39 (23%) were originally published in an European journal, most commonly in French, German, or Italian, indicating that these ideas were being discussed and shared internationally.

3.2. Types of pain

The retrieved files contained 131 direct references to pain, categorised during data extraction into 9 types, with 18 articles (14%) referencing more than one type of pain (Table 4). The most frequently reported pain categories were neuropathic and stump pain (as defined in Table 1), with 43 (33%) and 34 (26%) references, respectively. Stump pain was the only category to feature in every decade from the 1910s to the 1950s. Neuropathic pain was most commonly recorded during 1914 to 1918. Common terms for neuropathic pain (“neuralgia,” “neuritis,” and “causalgia”) were directly employed by the primary authors in 18 of articles (14%). Despite modern estimates that phantom limb sensation (as defined in Table 1) can affect up to 70% of amputees, no specific references to this condition were retrieved.1 Thirteen studies (13%) investigated causes and treatment of phantom limb pain, reaching a peak in the period 1946 to 1949.

Table 4 - Types of pain reported in retrieved texts. Pain type Total 1914–19 1920–29 1930–39 1940–45 1946–49 1950–59 Causalgia 11 5 5 1 — — — Hyperaesthesia 1 1 — — — — — Neuralgia 6 4 — 1 — 1 — Neuritis 1 1 — — — — — Neuropathic 43 28 9 4 1 1 — Phantom limb pain 13 — 2 1 2 5 3 Scar 3 1 — — 2 — — Sciatica 1 1 — — — — — Stump 34 16 2 5 5 3 3

Theories into potential causes of chronic postamputation pain were recorded and classified under 3 themes: “nociceptive,” “neuropathic,” or “psychological,” in accordance with the usage and concepts pertaining at the time of publication (Table 5). Although these 3 categories are not terms necessarily widely used by clinicians of the period, these descriptive terms were felt to be most appropriate by the authors. The most commonly reported theories were based around pain caused by a physiological mechanism often relating to treatment and rehabilitation. These included a poorly fitting prosthesis, amputation technique causing physical issues with the stump, scarring, or necrosed bone. Together, these accounted for 64 (50%) of all mechanistic descriptors. The second category, pain with a neuropathic origin, accounted for 54 references (42%) and also peaked during the First World War. The third category, “pain of a psychological origin,” accounted for 10 references (8%), and whilst this is an outdated concept for today's medicine, it was considered important to include due to the stigma that still surrounds chronic pain and the potential effect of untreated pain on a patient's mental health.

Table 5 - Mechanistic descriptors of postamputation pain in retrieved texts. Mechanistic descriptors Total 1914–19 1920–29 1930–39 1940–45 1946–49 1950–59 Nociceptive 64 39 7 6 7 4 1  Abscesses 3 1 — 1 — 1 —  Prosthetic limb fit 3 1 1 — — 1 —  Amputation technique 11 5 — 2 3 — 1  Treatment 7 3 1 1 2 — —  Bone spurs 2 2 — — — — —  Calluses 2 1 — — — 1 —  Foreign bodies 6 6 — — — — —  Jactitation 2 — — 1 — 1 —  Necrosis 2 2 — — — — —  New bone formation 1 1 — — — — —  Osteitis 3 2 — — 1 — —  Scar tissue 16 12 2 1 1 — —  Sepsis 2 — 2 — — — —  Sequestrum 2 2 — — — — —  Vascular 2 1 1 — — — — Neuropathic 54 39 8 2 2 1 2  Damage to nerve fibres 2 2 — — — — —  Infective inflammation 13 11 1 1 — — —  Lesions of nerve trunks 10 8 2 — — — —  Nerve concussion 3 3 — — — — —  Nerve regeneration 9 4 4 — — — 1  Nerves 2 1 — 1 — — —  Neuroma 15 10 1 — 2 1 1 Psychological 10 3 3 1 1 2 —  Psychical 9 3 3 1 1 1 —  Personality type 1 — — — — 1 —

The total number of participants described in the included reports was 9,326. Eleven studies included at least 100 participants; the largest reported 2000 patients. The mechanism of injury was specifically noted for only 168 patients. The most common cause reported was a gunshot wound, a term used during the First World War as inclusive of both projectile and blast injuries and which varied from a rifle bullet to shrapnel and fragments from artillery shells.

3.3. Interventions

Forty-three pain management interventions were described by authors (Table 6). The most frequently reported interventions were surgical or percutaneous needle-based therapies. Sixty-seven reports (66%) were made from a surgical perspective. Outcomes for interventions were reported in any manner in only 28 texts (28%). The most successful treatment reported was neuroma percussion, a “refreshingly simple method” based on the theory that repeated targeted pressure with a small mallet or bar on traumatised nerve endings would cause them to degenerate into fibrous tissue and render them into a painless state of “chronic concussion.”12 Articles referring to this treatment reported 2 positive and zero negative outcomes. Eight separate treatments reported a single negative outcome, although no treatments reported more than one.

Table 6 - Treatments for postamputation pain in retrieved texts. Treatment Total 1914–19 1920–29 1930–39 1940–45 1946–49 1950–59 Topical 56 42 2 1 3 3 5  Bath (unspecified) 2 2 — — — — —   Air bath 2 2 — — — — —   Wax bath 1 — — 1 — — —   Whirlpool bath 4 4 — — — — —  Compression 4 1 — — 1 2 —  Counter irritation 1 1 — — — — —  Electrical 14 13